A method and system for unloading grain from a wagon

By using real-time 3D modeling and virtual fence technology, the system automatically identifies the status of the freight car and controls the grain suction equipment, solving the problems of high reliance on manpower, low efficiency, and poor safety in existing technologies, and realizing automated and intelligent unloading of grain from the freight car.

CN120622118BActive Publication Date: 2025-10-17SINOGRAIN CHENGDU STORAGE RESEARCH INSTITUTE CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202511129501.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2025-08-13
Publication Date
2025-10-17
Estimated Expiration
2045-08-13

AI Technical Summary

Technical Problem

The existing use of grain suction equipment to unload grain from carriages is highly dependent on manpower, has low unloading efficiency, is prone to collision with the carriage structure, and has poor safety.

Method used

By acquiring point cloud data and image data of the inner wall of the carriage and the grain in real time, a real-time 3D model is generated, the grain pile is divided into zones, a virtual fence is constructed, the grain suction strategy is determined, and the automated operation of the grain suction equipment is controlled.

Benefits of technology

It realizes the automatic unloading of grain from the carriage, reduces the reliance on manpower, improves the unloading efficiency, avoids collision accidents and ensures safety.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120622118B_ABST
    Figure CN120622118B_ABST
Patent Text Reader

Abstract

The present application relates to the technical field of grain loading and unloading and conveying, and discloses a method and system for unloading grain from a carriage, aiming to solve the problem of high dependence on manpower in the prior art, and mainly comprising the following steps: acquiring point cloud data and image data of the inner wall of the carriage, the top reinforcing member of the carriage and the grain to be unloaded in real time, and generating a real-time three-dimensional model of the carriage; dividing the space of the carriage into multiple grain pile partitions according to the top reinforcing member of the carriage in the real-time three-dimensional model, determining the real-time grain surface height of each grain pile partition, and constructing a virtual fence in real time according to the inner wall of the carriage in the real-time three-dimensional model; determining the grain suction strategy of the grain suction equipment according to the real-time grain surface height of each grain pile partition in the virtual fence, and controlling the grain suction equipment to suck the grain in the carriage according to the grain suction strategy. The present application realizes the automation and intelligentization of the grain unloading operation, reduces the dependence on manpower for unloading grain from the carriage, improves the grain unloading efficiency, avoids equipment collision, and is suitable for various types of carriages.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of grain loading, unloading and conveying, and in particular to a method and system for unloading grain from a carriage. Background Art

[0002] In the existing grain transportation industry, grain suction equipment is a specialized device that utilizes the principle of negative pressure pneumatic conveying to unload bulk grain from trucks. In practice, when the suction head of the equipment is inserted into the grain pile in the truck compartment to be unloaded, the pressure differential between the inside and outside of the pipe causes grain particles to be drawn into the pipe by the high-speed airflow, forming a gas-solid two-phase flow. The grain is then transported through the conveying pipe to the designated location, where it is then separated from the gas by a discharger. The grain settles by gravity and is collected, while the air is purified and discharged. This technology enables mechanized, continuous, and enclosed operations for bulk grain unloading.

[0003] The existing technology relies heavily on manpower when using grain suction equipment to unload grain from trucks, resulting in at least the following problems: First, manual labor is required to continuously monitor the grain pile and equipment status, and the equipment needs to be frequently controlled, resulting in low grain unloading efficiency; second, manual labor is required to repeatedly check the height of the grain pile, and the equipment movement path relies on experience and has a large amount of invalid coverage; third, there is a risk of delayed response when manually monitoring the distance between the equipment and the carriage structure. When controlling the movement of the grain suction equipment, it is easy to collide with the carriage structure, resulting in poor safety. Summary of the Invention

[0004] The present invention aims to solve the problem of high manpower dependence in the existing method of unloading grain from a carriage using grain suction equipment, and proposes a method and system for unloading grain from a carriage.

[0005] The technical solution adopted by the present invention to solve the above technical problems is:

[0006] In a first aspect, the present invention provides a method for unloading grain from a carriage, the method comprising:

[0007] Acquire point cloud data and image data of the carriage inner wall, the carriage roof reinforcement structure, and the grain to be unloaded in real time, and generate a real-time three-dimensional model of the carriage based on the point cloud data and image data;

[0008] Dividing the carriage space into a plurality of grain pile sections based on the carriage roof reinforcement in the real-time three-dimensional model, determining the real-time grain surface height of each grain pile section, and constructing a virtual fence in real time based on the carriage inner wall in the real-time three-dimensional model to provide real-time motion constraints for the grain suction equipment;

[0009] In the virtual fence, the grain suction strategy of the grain suction equipment is determined according to the real-time grain surface height of each grain pile partition, and the grain suction equipment is controlled to suck the grain in the carriage according to the grain suction strategy.

[0010] Further, generating a real-time three-dimensional model of the carriage according to the point cloud data and the image data comprises:

[0011] The illumination component and the reflection component in the image data are dynamically separated by a dual-channel illumination perception network to generate a corresponding enhanced RGB image and a depth gradient map;

[0012] The depth gradient map is input into an improved edge-guided network, which uses deformable convolution to construct a multi-scale edge response field and generates an edge heat map through cross-layer feature fusion;

[0013] A dynamic channel pruning module is embedded in the YOLOv7-tiny architecture to construct a lightweight detection model, and the channel pruning threshold of the dynamic channel pruning module is dynamically adjusted according to the regional entropy value of the edge heat map;

[0014] The lightweight detection model is optimized, and the optimized lightweight detection model is used for target detection on the enhanced RGB image to obtain a target detection result. The optimization of the lightweight detection model includes introducing a bidirectional feature pyramid and a spatial attention mechanism into the detection head part of the lightweight detection model, and constructing a three-tuple geometric constraint of the carriage structure;

[0015] The edge heat map is projected to the point cloud space of the point cloud data for point cloud-image registration. For the occluded area, the point cloud is interpolated and completed according to the target detection result and based on the three-tuple geometric constraint to obtain a complete point cloud. The enhanced RGB image is mapped to the surface of the completed complete point cloud to generate a real-time three-dimensional model of the carriage.

[0016] Further, the optimized lightweight detection model is deployed using a mixed precision quantization strategy, which includes 8-bit fixed-point quantization of convolution layer weights, 4-bit adaptive dynamic quantization of activation layers, and hardware-aware operator fusion.

[0017] Further, the real-time grain surface height of each grain pile partition is the average grain surface height or the maximum grain surface height of the corresponding grain pile partition.

[0018] Further, the real-time grain surface height of each grain pile partition is the average grain surface height or the maximum grain surface height of the corresponding grain pile partition.

[0019] A target grain pile partition with the maximum real-time grain surface height is determined, a corresponding first grain suction path and a first moving speed are generated in the virtual fence according to the position of the target grain pile partition, and the grain suction equipment is controlled to suction the target grain pile partition according to the first grain suction path and the first moving speed;

[0020] When the real-time grain surface height of the target grain pile partition is less than the height threshold, the target grain pile partition with the maximum real-time grain surface height is re-determined, and the grain suction equipment is controlled to suck the grain in the re-determined target grain pile partition according to the first grain suction path and the first moving speed. In this way, the real-time grain surface height of all grain pile partitions is less than the height threshold.

[0021] Further, the control of the grain suction equipment to suck the grain in the target grain pile partition according to the first grain suction path and the first moving speed comprises:

[0022] After the suction head of the grain suction equipment is controlled to move to the target grain pile partition, the suction head is controlled to insert into the grain pile in the target grain pile partition and then to suck the grain while moving in the grain pile according to the first grain suction path and the first moving speed.

[0023] Further, the determination of the grain suction path of the grain suction equipment according to the real-time grain surface height of each grain pile partition further comprises:

[0024] When the real-time grain surface height of all grain pile partitions is less than the height threshold, the suction head of the grain suction equipment is controlled to suck the residual grain in the corner of the carriage and the residual grain in the vertical projection area of the top reinforcing member of the carriage according to a second moving speed, and the second moving speed is less than the first moving speed.

[0025] Further, the second moving speed is dynamically adjusted according to the surface area of the residual grain in the real-time three-dimensional model. The larger the surface area of the residual grain, the greater the second moving speed, and the smaller the surface area of the residual grain, the smaller the second moving speed.

[0026] In a second aspect, the present application provides a carriage grain unloading system for implementing the carriage grain unloading method as described in the first aspect, and the system comprises a radar module, a camera module, a controller and a grain suction equipment.

[0027] The radar module is used to obtain the point cloud data of the inner wall of the carriage, the top reinforcing member of the carriage and the grain to be unloaded in real time, and the camera module is used to obtain the image data of the inner wall of the carriage, the top reinforcing member of the carriage and the grain to be unloaded in real time.

[0028] The controller is used to generate a real-time three-dimensional model of the carriage according to the point cloud data and the image data; divide the space of the carriage into multiple grain pile partitions according to the top reinforcing member of the carriage in the real-time three-dimensional model, determine the real-time grain surface height of each grain pile partition, and construct a virtual fence for providing real-time motion constraints for the grain suction equipment according to the inner wall of the carriage in the real-time three-dimensional model; in the virtual fence, determine the grain suction strategy of the grain suction equipment according to the real-time grain surface height of each grain pile partition, and control the grain suction equipment to suck the grain in the carriage according to the grain suction strategy.

[0029] Further, the grain suction equipment comprises a moving door machine, a pneumatic conveying assembly and a grain conveying assembly which are electrically connected with the controller respectively.

[0030] The moving door machine comprises two front and rear door-shaped supports connected through a top cross beam and a middle cross beam, and a lifting platform slidingly arranged on the stand of the two door-shaped supports;

[0031] The pneumatic conveying assembly comprises a grain suction mechanism and an air suction mechanism which are arranged on the lifting platform and connected with each other;

[0032] The grain conveying assembly comprises a belt conveyor arranged on the lifting platform;

[0033] The grain suction mechanism comprises an extendable material pipe, a large-bend separator and a gas-solid separation pipe, two ends of the large-bend separator are respectively provided with an inlet and an outlet with downward opening, the inlet is connected with the extendable material pipe which can be vertically lifted, the outlet is rotatably connected with an inlet of the gas-solid separation pipe with a three-way structure, a downward inclined transition plate is arranged between the inlet and the outlet of the large-bend separator, so that the internal cavity of the large-bend separator has a shape with gradually increasing cross section from the inlet end to the outlet end, the corner part of the top of the large-bend separator opposite to the inlet and the outlet is provided with a circular arc chamfer, the outlet of the side of the gas-solid separation pipe is connected with the air suction mechanism, the outlet of the bottom is located above the inlet end of the belt conveyor, and a rotary unloading valve is arranged at the outlet of the bottom;

[0034] The extendable material pipe comprises a fixed pipe and an extendable pipe, the upper end of the fixed pipe is fixedly connected with the inlet of the large-bend separator, the upper end of the extendable pipe is sealingly and slidingly sleeved with the lower end of the fixed pipe, and the lower end of the extendable pipe is provided with a grain suction head.

[0035] The grain unloading method and system provided by the application can automatically identify the inner wall of the carriage, the top reinforcing member of the carriage and the grain to be unloaded, automatically divide the grain pile partitions according to the identification result, determine the real-time grain surface height, build a virtual fence, and finally control the grain suction equipment to automatically suck the grain in each grain pile partition according to the grain suction strategy corresponding to the real-time grain surface height in the virtual fence. The above process realizes automatic unloading of the grain in the carriage through automatic identification of the state of the carriage and the grain pile and automatic control of the unloading equipment, reduces the dependence on manpower, and can adapt to different vehicle structure differences to realize automatic and intelligent unloading operation. Specifically, through real-time grain surface height automatic partition identification, the optimal grain suction path is dynamically planned, the invalid travel is reduced, and the grain suction efficiency is improved; through automatic construction of the motion constraint boundary of the grain suction equipment by the virtual fence, collision accidents are avoided, and the safety is improved. BRIEF DESCRIPTION OF DRAWINGS

[0036] Figure 1 The flowchart of the grain unloading method in the embodiment is shown;

[0037] Figure 2 A three-dimensional structural schematic diagram of the grain suction equipment provided for the embodiment is shown in the figure;

[0038] Figure 3 A front view of the grain suction equipment provided for the embodiment is shown in the figure;

[0039] Figure 4 A structural schematic diagram of the grain unloading system of the carriage provided for the embodiment is shown in the figure;

[0040] Explanation of reference signs:

[0041] 1 - mobile door machine, 2 - grain feeding assembly, 4 - lifting platform, 5 - grain suction mechanism, 6 - air suction mechanism, 11 - door type support, 12 - top cross beam, 13 - middle cross beam, 51 - telescopic material pipe, 52 - large elbow separator, 53 - gas-solid separation pipe, 54 - rotary unloading valve, 111 - stand column, 511 - fixed pipe, 512 - telescopic pipe, 513 - grain suction head, 521 - feeding port, 522 - discharging port, 523 - transition plate. DETAILED DESCRIPTION

[0042] The technical solution of the present application is applicable to application scenarios in which the grain in the carriage needs to be unloaded, for example, the unloading of grains such as rice, corn, and wheat.

[0043] At present, the way of using the grain suction equipment to unload the carriage has a high degree of dependence on manpower, which results in a large number of invalid trips of the grain suction equipment, low grain suction efficiency, and easy collision with the carriage.

[0044] Therefore, the technical solution of the present application is proposed, in which, first, the point cloud data and image data of the inner wall of the carriage, the top reinforcing member of the carriage, and the grain are acquired in real time; thus, a real-time three-dimensional model of the carriage is generated; based on the real-time three-dimensional model, the top reinforcing member of the carriage is used to divide the carriage space into multiple grain pile partitions, and the real-time grain surface height of each grain pile partition is determined accordingly; at the same time, a virtual fence is constructed in real time according to the inner wall information in the real-time three-dimensional model to constrain the movement range of the grain suction equipment; within the virtual fence, the corresponding grain suction strategy is determined in combination with the real-time grain surface height of each grain pile partition; then, the grain suction equipment is controlled to perform the grain suction operation according to the grain suction strategy, thereby achieving the unloading of the carriage.

[0045] The above process can realize automatic unloading of grain for any type of carriage. Specifically, by fusing point cloud (spatial coordinates) and image (texture details) data in real time, the two types of data complement each other, and a real-time three-dimensional model of the carriage with millimeter-level accuracy can be generated in real time; a virtual fence is constructed according to the inner wall of the carriage in the real-time three-dimensional model, which can accurately reflect the current internal state of the carriage, provide real-time motion constraints for the grain suction equipment, and prevent the grain suction equipment from colliding with the inner wall of the carriage; according to the real-time grain surface height in the real-time three-dimensional model, the corresponding grain suction strategy is determined, which realizes efficient and stable batch suction in the grain sufficient stage, simplifies the control complexity, improves the operation efficiency and stability of grain unloading, and adjusts the grain suction strategy for residual grain to realize the complete removal of grain in the carriage, realizing efficient and comprehensive residual grain processing.

[0046] The technical solutions in the embodiments will be clearly and completely described below with reference to the drawings in the embodiments. Obviously, the described embodiments are only part of the embodiments of the present application, not all.

[0047] Figure 1 A flowchart of a carriage grain unloading method is shown, please refer to Figure 1 The method comprises the following steps:

[0048] Step 1: Real-time acquisition of point cloud data and image data of the inner wall of the carriage, the top reinforcing member of the carriage and the grain to be unloaded, and generation of a real-time three-dimensional model of the carriage according to the point cloud data and image data.

[0049] In actual application, the point cloud data of the inner wall of the carriage, the top reinforcing member of the carriage and the grain to be unloaded can be acquired by setting a radar module, and the image data of the inner wall of the carriage, the top reinforcing member of the carriage and the grain to be unloaded can be acquired in real time by a camera module. The top reinforcing member of the carriage is used to improve the strength of the top of the carriage, such as steel beams, steel wires, reinforcing ribs, etc.

[0050] After acquiring the point cloud data and image data, a real-time three-dimensional model of the carriage can be generated. In the grain unloading scene, the dust concentration on site can reach more than 5 mg / m³, the strong scattering effect causes the image contrast to decrease sharply, and the fine line target (diameter ≤8 mm) is almost overlapped with the carriage background in gray scale, which easily leads to image scattering noise and missing of point cloud data, and further makes the accuracy of the three-dimensional model poor.

[0051] In order to improve the accuracy of the three-dimensional model, in the present embodiment, the generation method of the real-time three-dimensional model specifically comprises steps 11 to 15.

[0052] Step 11: dynamically separating the illumination component and the reflection component in the image data by a dual-channel illumination perception network, to generate a corresponding enhanced RGB image and a depth gradient map.

[0053] This step is used for illumination separation and dust suppression, eliminating dust-induced illumination interference while preserving geometric cues. Specifically, the Dual-ILN dynamically separates the illumination component and the reflection component in the image data, where the reflection component is suppressed by an asymmetric convolution kernel group, and the enhanced RGB image and depth gradient map are output, providing low-noise visual input and geometric cues for subsequent steps. The depth gradient map contains scene edges but may be blurred.

[0054] The dual-channel illumination perception network includes: a globally deployed global illumination estimation subnet and a locally deployed local reflection recovery subnet, a physically constrained loss function constructed using the improved Retinex theory, and an asymmetric convolution kernel group (3x1 and 1x3 convolution kernel cascade) based on the dust particle scattering model.

[0055] Step 12: input the depth gradient map into the improved edge-guided network, which uses deformable convolution to construct a multi-scale edge response field and generates an edge heat map through cross-layer feature fusion.

[0056] This step is used to generate a high-confidence edge heat map to solve the problem of edge breakage and geometric distortion caused by dust, ensuring edge integrity and reasonableness. Specifically, the depth gradient map output by step 11 is input into the improved edge-guided network (Edge-Guided Net), which uses deformable convolution to construct a multi-scale edge response field and generates a high-confidence edge heat map through cross-layer feature fusion. The geometric cues are converted into robust edge priors to drive subsequent registration and detection.

[0057] The improved edge-guided network includes: a channel attention weighting mechanism for cross-layer feature fusion in a five-scale feature pyramid constructed by deformable convolution, and an adversarial training strategy based on edge curvature constraints.

[0058] Step 13: embed a dynamic channel pruning module in the YOLOv7-tiny architecture to construct a lightweight detection model, where the channel pruning threshold of the dynamic channel pruning module is dynamically adjusted according to the regional entropy value of the edge heat map.

[0059] This step is used to construct a lightweight detection model. Specifically, the dynamic channel pruning module generates a spatial weight matrix based on the regional entropy value of the edge heat map, constructs a channel importance scoring function, and uses Gumbel-Softmax to realize differentiable selection of different pruning rates, thereby realizing precise adaptive compression of detection network parameters, reducing the number of parameters, and improving real-time performance.

[0060] Step 14: The lightweight detection model is optimized, and the enhanced RGB image is subjected to target detection by using the optimized lightweight detection model to obtain a target detection result; the optimization of the lightweight detection model comprises: introducing a bidirectional feature pyramid and a spatial attention mechanism into the detection head part of the lightweight detection model, and constructing a three-tuple geometric constraint of the carriage structure.

[0061] This step is used to optimize the lightweight detection model. Specifically, the bidirectional feature pyramid (BiFPN) is used to fuse multi-scale features, the spatial attention mechanism is used to focus on key areas under dust interference, and the three-tuple geometric constraint is used to check the occluded target. Through the target detection by using the optimized lightweight detection model, anti-occlusion reasoning can be realized, and the false detection rate can be reduced.

[0062] Step 15: The edge heat map is projected to the point cloud space of the point cloud data for point cloud-image registration. For the occluded area, the point cloud is interpolated and completed according to the target detection result and based on the three-tuple geometric constraint to obtain a complete point cloud. The enhanced RGB image is mapped to the surface of the completed point cloud to generate a real-time three-dimensional model of the carriage.

[0063] This step is used to generate a real-time three-dimensional model. Specifically, since the point cloud data is missing due to occlusion, it needs to be fused with the image. In this embodiment, the edge heat map is projected to the point cloud space for registration, and the point cloud is interpolated and completed according to the three-tuple constraint for the occluded area. Finally, the enhanced RGB image is used to provide visual details without dust noise for texture mapping, and a real-time three-dimensional model of the carriage with texture is finally generated.

[0064] In this embodiment, the optimized lightweight detection model is deployed by using a mixed precision quantization strategy. The mixed precision quantization strategy comprises: the convolution layer weight is quantized by using 8-bit fixed-point quantization, the activation layer is quantized by using 4-bit adaptive dynamic quantization, and hardware-aware operator fusion is performed.

[0065] The convolution layer weight is quantized by using 8-bit fixed-point quantization, which can retain the key feature extraction capability of the detection model and ensure the small target recognition accuracy in the dust environment. The activation layer is quantized by using 4-bit adaptive dynamic quantization, which can cope with the brightness fluctuation caused by dust and greatly reduce data transfer and improve detection speed. The hardware-aware operator fusion can include merging Conv-BN-ReLU into a single calculation unit, optimizing 3x3 convolution calculation by using Winograd algorithm, designing a special memory reuse strategy for the target detection head, eliminating the inter-layer result storage after fusion, and optimizing data reuse according to the shared memory structure of the embedded GPU, thereby reducing the reasoning delay and the number of memory accesses and improving the real-time performance of the real-time three-dimensional model.

[0066] In the above scheme of constructing a real-time three-dimensional model of the carriage, the ambient light interference is eliminated through light separation, the complex light scene with high dust concentration under the unloading environment is adapted, the multi-scale edge response is realized through the edge heat map, the high-precision point cloud registration is supported, the registration failure caused by the blurred edge of the metal carriage is solved, the real-time requirement is met through dynamic pruning, the model integrity is ensured through geometric constraint completion, and the real-time performance of the real-time three-dimensional model is further improved through the mixed precision quantization strategy.

[0067] Step 2: dividing the carriage space into multiple grain pile partitions according to the carriage roof reinforcing members in the real-time three-dimensional model, determining the real-time grain surface height of each grain pile partition, and constructing a virtual fence for providing real-time motion constraints for the grain suction equipment according to the inner wall of the carriage in the real-time three-dimensional model.

[0068] In actual application, the carriage roof reinforcing members are determined from the real-time three-dimensional model of the carriage, and then the carriage is cut into multiple serial grain pile partitions along the direction of the members, for example, the area between two steel beams is divided into a grain pile partition. When calculating the real-time grain surface height of each grain pile partition, the point cloud corresponding to each grain pile partition is plane-fitted, and the average height or maximum height of the point set is taken as the real-time grain surface height. When constructing the virtual fence, the inner wall point cloud of the carriage is determined from the real-time three-dimensional model of the carriage, an isolation boundary is generated along the inner wall point cloud (appropriately inward by a certain safety distance), the isolation boundary is extended to the carriage bottom surface to form a vertical isolation surface, and multiple isolation surfaces form the virtual fence. In the process of dynamic updating of the real-time three-dimensional model of the carriage, the boundary of the virtual fence needs to be recalculated according to the updated model, and the real-time dynamic updating of the virtual fence is performed. In the updating and calculation process, an incremental updating strategy is adopted, only the affected fence area is recalculated, and the original boundary of the unaffected area remains unchanged, so that the updated virtual fence can accurately reflect the current state of the carriage interior, provide real-time motion constraints for the grain suction equipment, and prevent the grain suction equipment from colliding with the newly exposed inner wall of the carriage due to the failure to update the virtual fence in time.

[0069] Step 3: determining the grain suction strategy of the grain suction equipment according to the real-time grain surface height of each grain pile partition in the virtual fence, and controlling the grain suction equipment to suck the grain in the carriage according to the grain suction strategy.

[0070] Please refer to Figure 2 and Figure 3The grain suction equipment provided by the embodiment comprises a mobile door machine 1, a pneumatic conveying assembly and a grain conveying assembly 2 which are electrically connected with a controller respectively; the mobile door machine 1 comprises two front and rear door-shaped supports 11 which are connected through a top cross beam 12 and a middle cross beam 13, and a lifting platform 4 which is slidingly arranged on the vertical columns 111 of the two door-shaped supports 11; the pneumatic conveying assembly comprises a grain suction mechanism 5 and an air suction mechanism 6 which are arranged on the lifting platform 4 and are connected with each other; the grain conveying assembly 2 comprises a belt conveyor which is arranged on the lifting platform 4; the grain suction mechanism 5 comprises an extensible material pipe 51, a large-bend separator 52 and a gas-solid separation pipe 53, two ends of the large-bend separator 52 are respectively provided with a feeding port 521 and a discharging port 522 which are downwardly opened, the feeding port 521 is connected with the extensible material pipe 51 which can be vertically lifted and lowered, the discharging port 522 is rotatably connected with an inlet of the gas-solid separation pipe 53 which has a three-way structure, a transition plate 523 which is downwardly inclined is arranged between the feeding port 521 and the discharging port 522 of the large-bend separator 52, so that the internal cavity of the large-bend separator 52 has a shape which gradually increases from the cross section of the feeding end to the cross section of the discharging end, the corner parts of the top of the large-bend separator 52 which are opposite to the feeding port 521 and the discharging port 522 are provided with circular arc chamfers, an outlet of the side of the gas-solid separation pipe 53 is connected with the air suction mechanism 6, an outlet of the bottom is located above the feeding end of the belt conveyor, and a rotary discharging valve 54 is arranged at the outlet of the bottom; the extensible material pipe 51 comprises a fixed pipe 511 and an extensible pipe 512, the upper end of the fixed pipe 511 is fixedly connected with the feeding port 521 of the large-bend separator 52, the upper end of the extensible pipe 512 is sealingly and slidingly sleeved with the lower end of the fixed pipe 511, and the lower end of the extensible pipe 512 is provided with a grain suction head 513.

[0071] The mobile door machine 1 can be moved, the length and the width and height of the door-shaped supports 11 are designed according to the size of the commonly used truck carriage, and it is ensured that the truck can pass between the two vertical columns 111 of the door-shaped supports 11. The lifting platform 4 is used for controlling the height of the pneumatic conveying assembly and the grain conveying assembly 2, and the height is mainly adjusted according to the height of the carriage. The extensible material pipe 51 can be extended and retracted, so that the grain suction head 513 can be inserted into different depths of the grain.

[0072] Based on the above grain suction equipment, the controller can realize the positioning and grain suction of the suction head 513 in the carriage without blind area, automation and high precision by controlling the movement of the moving door machine 1, the height of the carriage by controlling the lifting platform 4, and the extension of the telescopic material pipe 51 and the rotation of the large-bend separator 52. In actual application, first, the carriage needs to be moved below the lifting platform 4 or the moving door machine 1 is controlled to move so that the carriage is below the lifting platform 4, and then the controller generates a corresponding control instruction set according to the suction path and movement speed of the suction head 513 and the current position coordinates and target position coordinates of the suction head 513 to control the movement of the moving door machine 1, the extension of the telescopic material pipe 51 and the rotation of the large-bend separator 52, so that the suction head 513 moves along the corresponding suction path.

[0073] In addition, the above grain suction equipment sets the large-bend separator 52 between the telescopic material pipe 51 and the gas-solid separation pipe 53. During the grain suction process, the gas-solid two-phase flow enters the large-bend separator 52 from the telescopic material pipe 51, and the channel cross section suddenly increases, the wind speed drops sharply, and the air carrying capacity of the material decreases. As a result, the material particles with large kinetic energy cannot make a sharp turn with the airflow due to inertia, thereby separating from the airflow and settling out under the action of gravity. This can not only reduce the breakage rate of grain, but also effectively separate grain and gas to avoid grain being sucked into the suction mechanism.

[0074] In this embodiment, the grain suction strategy includes a conventional grain suction strategy and a residual grain suction strategy, which dynamically adjusts the grain suction strategy according to the size of the real-time grain surface height and the preset height threshold of each grain pile partition.

[0075] In the conventional grain suction strategy, the target grain pile partition with the maximum real-time grain surface height is determined, a corresponding first suction path and a first movement speed are generated in the virtual fence according to the position of the target grain pile partition, and the grain suction equipment is controlled to suction the target grain pile partition according to the first suction path and the first movement speed. When the real-time grain surface height of the target grain pile partition is less than the height threshold, the target grain pile partition with the maximum real-time grain surface height is determined again, and the grain suction equipment is controlled to suction the newly determined target grain pile partition according to the first suction path and the first movement speed. In this way, the real-time grain surface height of all grain pile partitions is less than the height threshold.

[0076] Specifically, the conventional grain suction strategy controls the grain suction equipment to suction according to the principle of high layer first and low layer later, preferentially unloading high layer grain area. After the high layer grain pile is sucked, the low layer grain naturally collapses to fill the space, reducing the suction head empty suction probability, and avoiding the formation of a pit in the low layer, which leads to the risk of high layer grain suspension collapse. In addition, preferentially processing the high layer area with more serious dust raising can avoid concentrated dust raising in the later stage, and can also improve the accuracy of the real-time three-dimensional model. At the same time, preferentially unloading can reduce the risk of carriage inflation due to the greater pressure of high layer grain pile on the carriage side wall.

[0077] In the actual application of the conventional suction strategy, for each target grain pile partition, the suction head 513 of the suction equipment is first controlled to move to the target grain pile partition, and then the suction head 513 is controlled to insert into the grain pile in the target grain pile partition and perform suction, while the suction head 513 is controlled to move in the grain pile according to a first suction path and a first moving speed, to realize the standardized suction operation based on the fixed path and the fixed speed. The first suction path can be an S-shaped spiral path, and the height threshold and the first moving speed can be set according to actual conditions. The conventional suction strategy controls each grain pile partition according to the fixed path and constant speed control strategy in the sufficient grain stage to efficiently and stably batch suction, which simplifies the control complexity and significantly improves the operation efficiency and stability of the unloading.

[0078] When the real-time grain surface heights of all grain pile partitions are less than the height threshold, the residual grain suction strategy is executed. In the residual grain suction strategy, the moving path of the suction head 513 is adjusted according to the residual grain point cloud in the real-time three-dimensional model, and the suction head 513 is controlled to suck the residual grain in the corner of the carriage and the vertical projection area of the top reinforcing member of the carriage according to a second moving speed, which can cover all the corner and edge areas of the inner wall of the carriage, and realize the complete removal of the grain in the carriage.

[0079] The second moving speed is less than the first moving speed, because there is less residual grain in the carriage, and setting a smaller moving speed can achieve complete removal of the residual grain. In order to improve the accuracy of the second moving speed, the second moving speed is also dynamically adjusted according to the surface area of the residual grain in the real-time three-dimensional model. The larger the surface area of the residual grain, the larger the second moving speed, and the smaller the surface area of the residual grain, the smaller the second moving speed.

[0080] In actual application, the corresponding relationship between the surface area of the grain and the second moving speed can be constructed in advance, the surface area of the residual grain is determined through the residual grain point cloud in the real-time three-dimensional model, and the second moving speed is dynamically adjusted according to the surface area of the residual grain based on the corresponding relationship. Through dynamic adjustment of the second moving speed, the balance between the suction speed and the residual grain cleaning effect can be achieved, and the accuracy of the control of the suction equipment is improved.

[0081] In summary, the grain unloading method for the car provided in the embodiment can generate a real-time three-dimensional model of the car with millimeter-level precision in real time by fusing point cloud data and image data in real time, the two kinds of data complement each other, can generate a real-time three-dimensional model of the car with millimeter-level precision in real time; a virtual fence is constructed according to the inner wall of the car in the real-time three-dimensional model, which can accurately reflect the current internal state of the car, provide real-time motion constraints for the grain suction equipment, and prevent the grain suction equipment from colliding with the inner wall of the car; the corresponding grain suction strategy is determined according to the real-time grain surface height in the real-time three-dimensional model, efficient and stable batch suction is realized in the sufficient grain stage, the control complexity is simplified, the unloading efficiency and stability are improved, the residual grain is adjusted to realize the complete removal of the grain in the car, efficient and comprehensive residual grain treatment is realized, and automatic unloading can be realized for any type of car.

[0082] Based on the above technical solution, the embodiment further provides a car grain unloading system for realizing the car grain unloading method as described in the embodiment, please refer to Figure 4 , the system comprises a radar module, a camera module, a controller and a grain suction equipment;

[0083] The radar module is used to acquire point cloud data of the inner wall of the car, the top reinforcing member of the car and the grain to be unloaded in real time, and the camera module is used to acquire image data of the inner wall of the car, the top reinforcing member of the car and the grain to be unloaded in real time;

[0084] The controller is used to generate a real-time three-dimensional model of the car according to the point cloud data and the image data; the space of the car is divided into a plurality of grain pile partitions according to the top reinforcing member of the car in the real-time three-dimensional model, the real-time grain surface height of each grain pile partition is determined, and a virtual fence for providing real-time motion constraints for the grain suction equipment is constructed in real time according to the inner wall of the car in the real-time three-dimensional model; in the virtual fence, the grain suction strategy of the grain suction equipment is determined according to the real-time grain surface height of each grain pile partition, and the grain suction equipment is controlled to suck the grain in the car according to the grain suction strategy.

[0085] It can be understood that the car grain unloading system described in the embodiment is a system for realizing the car grain unloading method described in the embodiment, and for the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant part is described in the part of the method.

Claims

1. A method for unloading grain from a carriage, characterized in that: The method comprises: Acquire point cloud data and image data of the carriage inner wall, the carriage roof reinforcement structure, and the grain to be unloaded in real time, and generate a real-time three-dimensional model of the carriage based on the point cloud data and image data; Dividing the carriage space into a plurality of grain pile sections based on the carriage roof reinforcement in the real-time three-dimensional model, determining the real-time grain surface height of each grain pile section, and constructing a virtual fence in real time based on the carriage inner wall in the real-time three-dimensional model to provide real-time motion constraints for the grain suction equipment; Within the virtual fence, a grain suction strategy of the grain suction device is determined according to the real-time grain surface height of each grain pile partition, and the grain suction device is controlled to suck the grain in the carriage according to the grain suction strategy; The grain suction strategy of the grain suction equipment is determined based on the real-time grain surface height of each grain pile partition, including: Determine a target grain pile partition having the highest real-time grain surface height, generate a corresponding first grain suction path and a first moving speed within the virtual fence according to the position of the target grain pile partition, and control the grain suction device to suck grain from the target grain pile partition according to the first grain suction path and the first moving speed; When the real-time grain surface height of the target grain pile partition is less than the height threshold, the target grain pile partition with the largest real-time grain surface height is re-determined, and the grain suction device is controlled to suck grain from the re-determined target grain pile partition according to the first grain suction path and the first moving speed, and so on, until the real-time grain surface heights of all grain pile partitions are less than the height threshold; Controlling the grain suction equipment to suck grain from the target grain pile partition according to the first grain suction path and the first moving speed, including: After controlling the grain suction head of the grain suction device to move to the target grain pile partition, controlling the grain suction head to insert into the grain pile in the target grain pile partition to suck grain, and controlling the grain suction head to move in the grain pile according to a first grain suction path and a first moving speed; The grain suction path of the grain suction equipment is determined based on the real-time grain surface height of each grain pile partition, and also includes: When the real-time grain surface height of all grain pile partitions is less than the height threshold, the grain suction head of the grain suction equipment is controlled to suck the residual grain in the corners of the carriage and the vertical projection area of ​​the reinforcement structure on the top of the carriage at a second moving speed, and the second moving speed is less than the first moving speed.

2. The method for unloading grain from a carriage according to claim 1, characterized in that: Generating a real-time three-dimensional model of the carriage according to the point cloud data and the image data, including: Dynamically separate the illumination component and the reflection component in the image data through a dual-channel illumination perception network to generate corresponding enhanced RGB images and depth gradient maps; Inputting the depth gradient map into an improved edge-guided network, the improved edge-guided network uses deformable convolution to construct a multi-scale edge response field and generates an edge heat map through cross-layer feature fusion; A dynamic channel pruning module is embedded in the YOLOv7-tiny architecture to build a lightweight detection model. The channel pruning threshold of the dynamic channel pruning module is dynamically adjusted according to the regional entropy value of the edge heat map. Optimizing the lightweight detection model, and performing target detection on the enhanced RGB image using the optimized lightweight detection model to obtain target detection results; optimizing the lightweight detection model includes: introducing a bidirectional feature pyramid and a spatial attention mechanism into a detection head of the lightweight detection model, and constructing a triplet geometric constraint of the vehicle compartment structure; The edge heat map is projected into the point cloud space of the point cloud data for point cloud-image registration. For occluded areas, the point cloud is interpolated and completed based on the target detection results and triplet geometric constraints to obtain a complete point cloud. The enhanced RGB image is mapped to the surface of the completed point cloud to generate a real-time 3D model of the vehicle compartment.

3. The method for unloading grain from a carriage according to claim 2, characterized in that: The optimized lightweight detection model is deployed using a mixed-precision quantization strategy, which includes: 8-bit fixed-point quantization of convolutional layer weights, 4-bit adaptive dynamic quantization of activation layers, and hardware-aware operator fusion.

4. The method for unloading grain from a carriage according to claim 1, characterized in that: The real-time grain surface height of each grain pile partition is the average grain surface height or the maximum grain surface height of the corresponding grain pile partition.

5. The method for unloading grain from a carriage according to claim 1, wherein: The second moving speed is dynamically adjusted according to the surface area of ​​the remaining grain in the real-time three-dimensional model. The larger the surface area of ​​the remaining grain, the greater the second moving speed, and the smaller the surface area of ​​the remaining grain, the smaller the second moving speed.

6. A carriage grain unloading system, characterized in that: Used to implement the method for unloading grain from a carriage as claimed in any one of claims 1 to 5, the system comprises a radar module, a camera module, a controller and a grain suction device; The radar module is used to obtain point cloud data of the inner wall of the carriage, the reinforcement structure on the top of the carriage, and the grain to be unloaded in real time, and the camera module is used to obtain image data of the inner wall of the carriage, the reinforcement structure on the top of the carriage, and the grain to be unloaded in real time; The controller is used to generate a real-time three-dimensional model of the carriage based on the point cloud data and image data; The carriage space is divided into multiple grain pile partitions according to the carriage roof reinforcement structure in the real-time three-dimensional model, the real-time grain surface height of each grain pile partition is determined, and a virtual fence for providing real-time motion constraints for the grain suction equipment is constructed in real time according to the inner wall of the carriage in the real-time three-dimensional model; within the virtual fence, the grain suction strategy of the grain suction equipment is determined according to the real-time grain surface height of each grain pile partition, and the grain suction equipment is controlled to suck the grain in the carriage according to the grain suction strategy.

7. The carriage grain unloading system according to claim 6, characterized in that: The grain suction equipment includes a movable door machine, a pneumatic conveying component and a grain delivery component which are electrically connected to the controller respectively; The mobile gantry crane comprises two front and rear gantry supports connected by a top crossbeam and a middle crossbeam, and a lifting platform slidably arranged on the columns of the two gantry supports; The pneumatic conveying assembly includes a grain suction mechanism and an air extraction mechanism that are arranged on a lifting platform and are connected to each other; The food delivery component includes a belt conveyor arranged on a lifting platform; The grain suction mechanism includes a telescopic material pipe, a large elbow separator and an air-solid separation pipe, two ends of the large elbow separator are respectively provided with a feed port and a discharge port opening downwards, the feed port is connected to the telescopic material pipe which can be lifted vertically, and the discharge port is rotatably connected to the inlet above the air-solid separation pipe of the three-way structure, a downward-sloping transition plate is provided between the feed port and the discharge port of the large elbow separator, so that the internal cavity has a shape in which the cross-section gradually increases from the feed end to the discharge end, and a circular arc chamfer is provided at the corner of the top of the large elbow separator facing the feed port and the discharge port, the outlet on the side of the air-solid separation pipe is connected to the exhaust mechanism, the outlet at the bottom is located above the feed end of the belt conveyor, and a rotary discharge valve is provided at the bottom outlet; The telescopic material pipe includes a fixed pipe and a telescopic pipe. The upper end of the fixed pipe is fixedly connected to the feed port of the large elbow separator. The upper end of the telescopic pipe is sealed and slidably sleeved with the lower end of the fixed pipe. The lower end of the telescopic pipe is provided with a grain suction head.

Citation Information

Patent Citations

  • Biasing map matched trajectories based on planned route information

    US10533862B1

  • Methods and systems for managing large asset fleets through a virtual reality interface

    US20170262786A1