CN-121982352-A - Renewable resource sorting control method and system based on intelligent cabinet

CN121982352ACN 121982352 ACN121982352 ACN 121982352ACN-121982352-A

Abstract

The invention provides a renewable resource sorting control method and system based on an intelligent cabinet, wherein the method comprises the steps of obtaining a segmentation mask and object category information by carrying out image recognition on renewable resource objects put in the intelligent cabinet; the method comprises the steps of calculating the grabbing position information of a manipulator based on a segmentation mask, analyzing the stacking relation among objects by combining depth information and generating a grabbing priority sequence, and generating a manipulator control instruction to realize sequential grabbing and classified storage. According to the invention, by introducing a spatial relationship analysis and priority ordering mechanism, the problem of collision of multi-object grabbing in a disordered environment is effectively solved, the accuracy and efficiency of automatic sorting of the renewable resources are remarkably improved, and a reliable technical scheme is provided for intelligent recovery equipment.

Inventors

ZHAO XIAOFENG

Assignees

深圳能源环保股份有限公司

Dates

Publication Date: 20260505
Application Date: 20251129

Claims (8)

1. The renewable resource sorting control method based on the intelligent cabinet is characterized by comprising the following steps of: Carrying out image recognition on the renewable resource objects put in the intelligent cabinet to obtain a segmentation mask and object type information of each renewable resource object; Performing grabbing coordinate calculation processing on each renewable resource object according to the segmentation mask to obtain grabbing position information of the manipulator; Performing spatial relationship analysis processing on each renewable resource object according to the segmentation mask and the depth information to obtain a stacking relationship among the renewable resource objects, and generating a grabbing priority sequence according to the stacking relationship; And driving a manipulator in the intelligent cabinet to sequentially grab empty boxes according to the object category information, the grabbing position information and the grabbing priority sequence, marking renewable resource objects according to the categories, and completing unified storage.
2. The method for controlling sorting of renewable resources based on intelligent cabinets according to claim 1, wherein the performing image recognition on the renewable resource objects put in the intelligent cabinets to obtain the segmentation mask and object class information of each renewable resource object comprises: acquiring original image data of a renewable resource object in an intelligent cabinet through an RGB-D camera in the intelligent cabinet, and inputting the original image data into a pre-trained R-CNN neural network for feature extraction processing to obtain a multi-scale feature map; generating candidate areas of the multi-scale feature images through an area proposal network to obtain object candidate frames; Performing feature alignment processing through RoIAlign algorithm according to the object candidate frame to obtain feature mapping with fixed size; And respectively carrying out classification branch processing and mask branch processing on the feature map to obtain a segmentation mask and object category information of each renewable resource object.
3. The method for controlling sorting of renewable resources based on intelligent cabinets according to claim 1, wherein the performing a grabbing coordinate calculation process on each renewable resource object according to the segmentation mask to obtain grabbing position information of the manipulator includes: Performing binarization processing on the segmentation mask to obtain binarization areas of each renewable resource object; Extracting boundary pixels of the binarization area through a contour detection algorithm to obtain a boundary point set of an object contour; Calculating the minimum circumscribed rectangle of the renewable resource object according to the boundary point set, and calculating the length axis direction of the minimum circumscribed rectangle through a principal component analysis algorithm to obtain the main axis direction and the minor axis direction of the object; and generating a fixed length line segment perpendicular to the minor axis direction as a grabbing area mark by taking the intersection point of the minor axis direction and the boundary point set of the object profile as a reference, so as to obtain grabbing position information of the manipulator.
4. The intelligent cabinet-based renewable resource sorting control method according to claim 1, wherein the stacking relationship includes shielding status information between objects of the renewable resources; The step of carrying out spatial relationship analysis on the renewable resource objects according to the segmentation mask and the depth information to obtain a stacking relationship among the renewable resource objects, and the step of generating a grabbing priority sequence according to the stacking relationship comprises the following steps: calculating the minimum distance and the midpoint position between the outlines of the renewable resource objects based on the segmentation mask; determining shielding state information among the renewable resource objects according to the minimum distance, the midpoint position and the depth information; and constructing a relation matrix based on the shielding state information, and generating a collision-free grabbing priority sequence according to the relation matrix.
5. The intelligent cabinet-based renewable resource sorting control method according to claim 4, wherein determining the shielding status information between the renewable resource objects according to the minimum distance, midpoint position, and depth information comprises: comparing and judging the minimum distance with a preset distance threshold, and if the minimum distance is smaller than the preset distance threshold, determining that two corresponding renewable resource objects have a space adjacent relation; Constructing a square patch area with a side length of a preset pixel size as a depth sampling area by taking the midpoint position as a geometric center; Calculating the average depth value of the corresponding pixels of each renewable resource object in the depth sampling area, marking an object with a larger average depth value as an upper object, marking an object with a smaller average depth value as a lower object, and obtaining the upper-lower layer position relation among the renewable resource objects; and generating shielding state identifiers among the regenerated resource objects based on the spatial adjacent relationship and the upper-lower layer position relationship.
6. The intelligent cabinet-based renewable resource sorting control method according to claim 4, wherein the constructing a relationship matrix based on the shielding state information, and generating a collision-free grabbing priority sequence according to the relationship matrix comprises: According to the shielding state information, carrying out shielding relation value distribution processing on each pair of renewable resource objects to obtain a relation matrix, wherein if an x object in the renewable resource objects is positioned on the upper layer of a y object and has shielding relation, the x row and the y column of the relation matrix are assigned with 1, the y row and the x column are assigned with-1, and if no shielding relation exists between the two objects, the corresponding position is assigned with 0; Performing row scanning processing on the relation matrix, identifying row indexes of which row elements only contain 1 and 0 and do not contain-1 in the relation matrix, and marking a renewable resource object corresponding to the row indexes as a current grippable object; setting all elements corresponding to the row index and the column index of the current grippable object in the relation matrix to zero, thereby obtaining an updated relation matrix; Repeatedly executing line scanning processing and zero setting processing on the updated relation matrix until all elements in the relation matrix are zero; And generating a collision-free grabbing priority sequence according to the sequence of the objects of the renewable resources marked as the current grabbed objects.
7. The intelligent cabinet-based renewable resource sorting control method according to claim 1, wherein driving a manipulator in the intelligent cabinet to sequentially grasp empty boxes according to the object category information, the grasping position information and the grasping priority sequence, identifying renewable resource objects by category and completing unified storage comprises: acquiring unique two-dimensional code identification information of a unified storage container; Performing grabbing task sequencing processing on each renewable resource object based on the grabbing priority sequence, and generating a mechanical arm movement track parameter by combining the grabbing position information and the unified storage container position coordinates; Generating a manipulator control instruction based on the motion trail parameters, driving a manipulator to sequentially grab each renewable resource object from the corresponding position of the grabbing position information according to the grabbing priority sequence, and uniformly throwing the renewable resource objects of different categories into the same container; And establishing a binding relation between the unique two-dimensional code identification of the container and object type information of each renewable resource in the container, and acquiring the object type information in the container by scanning the two-dimensional code of the container for processing in the subsequent process.
8. Renewable resource sorting control system based on intelligent cabinet, characterized by comprising: The image recognition unit is used for carrying out image recognition on the renewable resource objects put in the intelligent cabinet to obtain the segmentation mask and object type information of each renewable resource object; the coordinate calculation unit is used for carrying out grabbing coordinate calculation processing on each renewable resource object according to the segmentation mask to obtain grabbing position information of the manipulator; The relation analysis unit is used for carrying out space relation analysis processing on each renewable resource object according to the segmentation mask and the depth information to obtain a stacking relation among the renewable resource objects, and generating a grabbing priority sequence according to the stacking relation; And the control execution unit is used for generating a manipulator control instruction according to the object category information, the grabbing position information and the grabbing priority sequence, and driving a manipulator in the intelligent cabinet to grab the renewable resource objects in sequence and finish classified storage.

Description

Renewable resource sorting control method and system based on intelligent cabinet Technical Field The invention relates to the technical field of image recognition, in particular to a renewable resource sorting control method and system based on an intelligent cabinet. Background With the enhancement of environmental awareness and the development of recycling economy, the recycling of renewable resources has become an important point of social concern. Traditional renewable resource sorting mainly relies on manual operation, and has the problems of low sorting efficiency, high labor cost, poor sorting accuracy and the like. Although some automatic sorting devices exist in the prior art, most of the automatic sorting devices adopt a single sensor identification mode, and it is difficult to accurately identify and position various renewable resource objects in a complex environment. The existing robot grabbing technology often lacks effective space relation analysis and grabbing sequence planning capability when dealing with a plurality of objects with disordered stacking, is easy to cause object collision or grabbing failure in the grabbing process, and cannot meet the actual requirement of efficient sorting of renewable resources. Particularly in the intelligent recovery equipment, how to realize accurate identification, reasonable sequencing and accurate grabbing of the randomly thrown renewable resources and improve the storage capacity of a unit area is still a technical problem to be solved. Disclosure of Invention The invention mainly aims to solve the technical problems of low sorting efficiency and high grabbing failure rate caused by the lack of effective multi-object space relation analysis and grabbing sequence planning in the existing renewable resource sorting technology; the first aspect of the invention provides a renewable resource sorting control method based on an intelligent cabinet, which comprises the following steps: Carrying out image recognition on the renewable resource objects put in the intelligent cabinet to obtain a segmentation mask and object type information of each renewable resource object; Performing grabbing coordinate calculation processing on each renewable resource object according to the segmentation mask to obtain grabbing position information of the manipulator; Performing spatial relationship analysis processing on each renewable resource object according to the segmentation mask and the depth information to obtain a stacking relationship among the renewable resource objects, and generating a grabbing priority sequence according to the stacking relationship; and generating a manipulator control instruction according to the object category information, the grabbing position information and the grabbing priority sequence, and driving a manipulator in the intelligent cabinet to grab the renewable resource objects in sequence and finish classified storage. As a preferred solution, the performing image recognition on the renewable resource objects put in the intelligent cabinet to obtain the segmentation mask and object category information of each renewable resource object includes: acquiring original image data of a renewable resource object in an intelligent cabinet through an RGB-D camera in the intelligent cabinet, and inputting the original image data into a pre-trained R-CNN neural network for feature extraction processing to obtain a multi-scale feature map; generating candidate areas of the multi-scale feature images through an area proposal network to obtain object candidate frames; Performing feature alignment processing through RoIAlign algorithm according to the object candidate frame to obtain feature mapping with fixed size; And respectively carrying out classification branch processing and mask branch processing on the feature map to obtain a segmentation mask and object category information of each renewable resource object. As a preferable solution, the performing, according to the segmentation mask, a grabbing coordinate calculation process on each renewable resource object, and obtaining grabbing position information of the manipulator includes: Performing binarization processing on the segmentation mask to obtain binarization areas of each renewable resource object; Extracting boundary pixels of the binarization area through a contour detection algorithm to obtain a boundary point set of an object contour; Calculating the minimum circumscribed rectangle of the renewable resource object according to the boundary point set, and calculating the length axis direction of the minimum circumscribed rectangle through a principal component analysis algorithm to obtain the main axis direction and the minor axis direction of the object; and generating a fixed length line segment perpendicular to the minor axis direction as a grabbing area mark by taking the intersection point of the minor axis direction and the boundary point set of the object profile as a refe