CN-122009813-A - Multi-dimensional feature fusion-based non-destructive oriented delivery system and method for fritillary bulb

CN122009813ACN 122009813 ACN122009813 ACN 122009813ACN-122009813-A

Abstract

The application provides a non-destructive directional delivery system and method for fritillary bulb based on multidimensional feature fusion, and belongs to the field of automatic processing of traditional Chinese medicinal materials. The application aims to solve the problems that medicinal materials such as fritillary bulb and the like are easy to damage in automatic feeding and difficult to identify due to surface dirt. The system comprises a feeding unit, a visual detection unit and an orientation placing unit. The automatic pressure relief device is characterized in that feeding speed is dynamically adjusted along with material density, double-layer contour matching is adopted in visual detection, micro-overturning secondary detection is triggered if confidence coefficient is lower than a threshold value, clamping force is monitored in real time by a flexible clamping jaw, and the medicinal materials are automatically relieved and protected once the clamping force exceeds the limit. According to the application, through the depth cooperative closed loop of vision and force sense, the efficient nondestructive orientation of fragile and irregular medicinal materials is realized, and the material utilization rate and the production efficiency are greatly improved.

Inventors

YU GUOHONG
ZHENG HANG
XUE XIANGLEI
REN NING
YE YUNXIANG
LIU TIAN
YAN YU

Assignees

浙江省农业科学院

Dates

Publication Date: 20260512
Application Date: 20260414

Claims (10)

1. The non-destructive oriented delivery system of fritillary bulb based on multidimensional feature fusion is characterized by comprising: The feeding unit is used for conveying medicinal materials to be treated; The visual detection unit is arranged at the downstream of the feeding unit and is used for collecting image information of medicinal materials and outputting posture data and posture identification confidence of the medicinal materials based on an image processing algorithm; the directional placing unit comprises a mechanical arm and a flexible clamping mechanism arranged at the tail end of the mechanical arm, wherein a pressure detection module is integrated in the flexible clamping mechanism and used for feeding back clamping force in real time; The main control unit is respectively in communication connection with the feeding unit, the visual detection unit and the directional placement unit; The master control unit is configured to execute the following cooperative control logic: When the gesture recognition confidence is lower than a preset validity threshold, controlling the directional placement unit to execute micro-overturning action to change the gesture of the medicinal material, and triggering the visual detection unit to acquire and recognize a secondary image; When the clamping force fed back by the pressure detection module exceeds a preset safety pressure threshold, the flexible clamping mechanism is controlled to execute active pressure relief operation until the clamping force is reduced to be within a safety range.
2. The non-destructive directional delivery system of Bulbus Fritillariae Thunbergii of claim 1, wherein the visual detection unit calculates the pose data using a two-layer contour matching algorithm comprising: the rough matching layer is used for carrying out fan-shaped sampling and principal component analysis based on geometric features of the contours of the medicinal materials and determining the main axis direction and the approximate direction of the medicinal materials; The fine matching layer is used for extracting texture trend characteristics and specific anatomical structure characteristics of the surface of the medicinal material on the basis of coarse matching, and judging the front and back surfaces and the rotation angle of the medicinal material through multi-scale texture analysis; the gesture recognition confidence is calculated according to the feature matching degree weighting fusion of the coarse matching layer and the fine matching layer.
3. The non-destructive directional delivery system of thunberg fritillary bulb according to claim 2, wherein the specific anatomical feature comprises a depressed umbilical region feature of the medicinal material, the visual detection unit comprises two groups of camera assemblies symmetrically arranged at a preset included angle and an annular diffuse reflection light source with adjustable brightness arranged around the camera assemblies, the two groups of camera assemblies are respectively used for capturing a top view image and a side view image of the medicinal material, and a three-dimensional posture model of the medicinal material is established through stereoscopic vision fusion calculation.
4. The non-destructive directional delivery system of fritillary bulb according to claim 1, wherein the specific execution logic of the micro-flipping action is: The method comprises the steps of controlling the mechanical arm to drive the flexible clamping mechanism to grab medicinal materials with a preset low clamping force lower than a normal grabbing force, controlling the tail end of the mechanical arm to execute space overturning movement with an angle smaller than or equal to 30 degrees, and adjusting clamping opening and closing degrees in real time according to feedback of the pressure detection module in the overturning process so as to prevent the medicinal materials from sliding down.
5. The non-destructive directional delivery system of fritillary bulb according to claim 1, wherein the flexible clamping mechanism comprises three circumferentially distributed silicone fingers, the inner side surfaces of the silicone fingers are provided with anti-slip micro-textures, and the active pressure relief operation is specifically as follows: when the clamping force exceeds 8N, the pneumatic proportional regulating valve is controlled to discharge 50% of air pressure in the air passage, so that the clamping force is instantaneously reduced.
6. The non-destructive directional delivery system of Bulbus Fritillariae Thunbergii of claim 5, wherein the feeding unit comprises a roller feeding mechanism and a conveying track connected thereto; the visual detection unit is further configured to count the number of medicinal materials passing through the detection area in real time in unit time to generate density data; The main control unit dynamically adjusts the conveying speed of the feeding unit according to the density data, wherein the conveying speed is reduced when the density data is higher than a preset dense threshold value, and the conveying speed is improved when the density data is lower than a preset sparse threshold value.
7. The non-destructive directed delivery system of fritillary bulb according to claim 1, further comprising a sample increment learning module for recording correct pose image data for a human intervention and dynamically updating a feature template library of the visual inspection unit using the image data as a training sample.
8. The non-destructive directional delivery system of fritillary bulb according to claim 1, wherein the silicone finger hardness of the flexible clamping mechanism is 30-40 degrees shore, and the width of the conveying track is customized to 35mm in accordance with the diameter range of the fritillary bulb.
9. A delivery method using the non-destructive directional delivery system of fritillary bulb according to any one of claims 1 to 8, comprising the steps of: s1, conveying medicinal materials to a visual detection area through a feeding unit; s2, acquiring a medicinal material image by using a visual detection unit, and calculating a gesture recognition confidence; S3, judging whether the confidence coefficient is larger than or equal to a preset validity threshold value, if so, entering a step S5, otherwise, controlling the directional placement unit to grasp the medicinal materials and slightly turn over, and then returning to the step S2 for secondary detection; S4, if the confidence coefficient of the secondary detection is still lower than the validity threshold, sending an alarm signal to prompt manual intervention; s5, planning a motion path by the main control unit according to the finally confirmed gesture data, and controlling the directional placement unit to grasp medicinal materials; S6, monitoring clamping force in real time in the grabbing and carrying processes, and immediately executing active pressure relief operation if the clamping force exceeds a safety pressure threshold; And S7, placing the medicinal materials at a target position according to a specified gesture, and triggering the next round of feeding by a feedback signal.
10. The delivery method according to claim 9, wherein the step of calculating the gesture recognition confidence in step S2 specifically includes: performing self-adaptive median filtering and local brightness equalization on the acquired image; Extracting outline radian characteristics of the traditional Chinese medicinal materials in the image, and performing rough matching with a standard elliptical template; extracting scale textures and umbilical features of the image Chinese medicinal materials, and performing fine matching with a pre-stored texture feature library; And obtaining the comprehensive confidence coefficient by weighting calculation according to the similarity scores of the coarse matching and the fine matching.

Description

Multi-dimensional feature fusion-based non-destructive oriented delivery system and method for fritillary bulb Technical Field The application relates to the technical field of primary processing and automatic equipment of traditional Chinese medicine production places, in particular to a non-destructive directional delivery system and method of fritillary bulb based on multidimensional feature fusion. Background The fritillary bulb is a common traditional Chinese medicine for clearing heat and resolving phlegm, the bulb is usually flat with the thickness of 5-8mm and the diameter of 15-35mm, and the surface is often provided with soil residues or natural irregular textures. In the processing process of the traditional Chinese medicine decoction pieces, orientation and feeding before slicing are key procedures. The existing pretreatment of the fritillary bulb mainly has two modes, namely manual throwing, manual front and back recognition and direction adjustment of workers are needed, the labor intensity is high, the production efficiency is low, and the requirement of large-scale production is difficult to meet. Secondly, general automatic feeding equipment (such as a vibrating plate or a general manipulator) is adopted, but the specific material of the fritillary bulb has obvious defects: 1. The physical damage rate is high, the epidermis of the thunberg fritillary bulb is easy to be damaged (the damage threshold is only about 5N), and the rigid grabbing or constant force clamping of general equipment is easy to cause the clamping or crushing of medicinal materials, so that the grade and selling price of the medicinal materials are influenced. 2. The vision recognition robustness is poor, the feature blurring is usually caused by mud dipping or light reflection on the surface of the thunberg fritillary bulb, the general vision algorithm usually directly judges the materials as waste materials to cause great material waste, or the grabbing gesture is wrong due to the fact that flat features have blind areas under a single visual angle. Therefore, there is a need for an intelligent device that can integrate visual and force feedback, protect fragile materials, and solve the recognition problem by actively adjusting the posture. Disclosure of Invention The embodiment of the application provides a multi-dimensional feature fusion-based non-destructive oriented delivery system and method for fritillary bulb, which aim at the problems that the prior art mainly relies on low-efficiency manual operation, and general automatic equipment lacks an adaptive protection and fault tolerance mechanism for fragile, flat and surface dirty medicinal materials, so that the breakage rate of the medicinal materials is high, the material error rejection rate caused by recognition interference is high, and the like. The core technology of the invention mainly builds a closed loop system of 'vision-force sense depth coordination', solves the difficult problem of identifying the complex surface through 'double-layer profile matching' and 'confidence-driven micro-overturning secondary detection', and realizes nondestructive grabbing of fragile medicinal materials by utilizing a 'pressure threshold automatic pressure release' mechanism. In a first aspect, the application provides a non-destructive oriented delivery system of fritillary bulb based on multidimensional feature fusion, comprising: The feeding unit is used for conveying medicinal materials to be treated; The visual detection unit is arranged at the downstream of the feeding unit and is used for collecting image information of the medicinal materials and outputting posture data and posture identification confidence of the medicinal materials based on an image processing algorithm; The directional placing unit comprises a mechanical arm and a flexible clamping mechanism arranged at the tail end of the mechanical arm, wherein a pressure detection module is integrated in the flexible clamping mechanism and used for feeding back clamping force in real time; the main control unit is respectively in communication connection with the feeding unit, the visual detection unit and the directional placement unit; The master control unit is configured to execute the following cooperative control logic: When the gesture recognition confidence is lower than a preset validity threshold, controlling the directional placement unit to execute micro-overturning action to change the gesture of the medicinal material, and triggering the visual detection unit to acquire and recognize a secondary image; when the clamping force fed back by the pressure detection module exceeds a preset safety pressure threshold, the flexible clamping mechanism is controlled to execute active pressure relief operation until the clamping force is reduced to be within a safety range. Further, the visual detection unit calculates the gesture data using a double-layer contour matching algorithm, the double-layer conto