Search

CN-122023866-A - Cotton plant topping system and method based on image sensing and intelligent control

CN122023866ACN 122023866 ACN122023866 ACN 122023866ACN-122023866-A

Abstract

The invention discloses a cotton plant topping system and method based on image sensing and intelligent control, which are used for realizing automatic identification, spatial positioning and accurate shearing of cotton plant terminal buds by integrating machine vision, gesture sensing and intelligent control technologies. The system collects image and attitude information through an industrial camera, an Inertial Measurement Unit (IMU) and a vehicle speed sensor, and performs image processing and decision control through an industrial control computer so as to realize topping path planning and intelligent execution. Meanwhile, the system can detect and predict the disease spots and insect damage areas of cotton plant leaves, and assist in realizing cotton field health state monitoring. The invention improves the automation and intelligent level of cotton field operation, and is suitable for the fields of accurate management of cotton and intelligent agricultural machinery equipment.

Inventors

  • NI CHAO
  • Li Ningran
  • XUE SHENG
  • ZHU TINGTING
  • CHEN CHUMIN

Assignees

  • 南京林业大学

Dates

Publication Date
20260512
Application Date
20251215

Claims (10)

  1. 1. The cotton plant topping system based on image sensing and intelligent control is characterized by comprising a machine body (1), an image and gesture sensing module (2), a topping execution module (3) and an image processing and intelligent control module (4), wherein the machine body (1) comprises a moving platform, a metal frame (11) and travelling wheels (12) which are used for bearing all functional modules and moving in a cotton field row, the image and gesture sensing module (2) comprises an industrial RGB camera (21), an inertial measurement unit IMU (22) and a vehicle speed sensor (23) which are arranged above and in front of a machine body shell, and are used for synchronously acquiring cotton plant canopy images and vehicle motion gesture information, the topping execution module (3) comprises a lifting module (31), a guide sliding rod (32), a shearing driving assembly and a shearing end actuator (33), the lifting module (31) and the guide sliding rod (32) are used for driving and restraining the shearing end actuator (33) to move in the vertical direction, the shearing end actuator (33) is used for shearing cotton plant top buds, and the image processing and intelligent control module (4) comprises an industrial control computer, and the industrial control unit IMU (21), the inertial measurement unit (23) and the inertial sensor (23) are respectively connected with the industrial control module (21) and the vehicle topping execution module (23).
  2. 2. The cotton plant topping system based on image sensing and intelligent control according to claim 1, wherein the travelling wheel (12) is made of wear-resistant rubber material, and an anti-skid pattern is arranged on the surface of the travelling wheel (12).
  3. 3. The cotton plant topping system based on image sensing and intelligent control according to claim 1, wherein the shearing end effector (33) comprises a rotary blade assembly, an annular limiting cover is arranged on the outer side of the rotary blade assembly, and the rotary blade assembly is made of corrosion-resistant stainless steel materials.
  4. 4. The cotton plant topping method based on image sensing and intelligent control is characterized by comprising the following steps of: S1, synchronously acquiring cotton plant canopy images, vehicle speed and attitude information through an industrial RGB camera, a vehicle speed sensor and an inertial measurement unit IMU, performing camera calibration and distortion correction on the acquired images, and performing image stabilization, deblurring and illumination normalization processing by combining the vehicle speed and attitude information to generate an enhanced image dataset containing a cotton plant top interest region; S2, in the enhanced image dataset, performing line and zone detection and canopy segmentation on cotton plant canopy, defining a working line area and obtaining canopy mask, constructing a top end interest area in the upper area of the canopy, inputting the top end interest area into a key point identification network, outputting pixel coordinates and confidence coefficient of a top bud key point, and mapping the top bud key point from an image plane to a machine body coordinate system by combining camera internal parameters and camera-machine body external parameters to obtain a top bud space position and corresponding confidence coefficient; s3, calculating color enhancement indication quantity in an effective search domain limited by the enhanced image and the canopy mask, carrying out region division and feature extraction on the search domain, acquiring candidate pest patches based on an abnormal degree function, inputting a sub-image corresponding to the candidate patches into a pest detection network to output a pest category, a boundary frame and confidence, calculating a pest risk score by combining the abnormal degree and the confidence, and completing conversion from pixel coordinates to ground coordinates and generating pest risk report data when the risk score exceeds a preset threshold value; S4, judging operation permission based on the terminal bud space position, the confidence coefficient, the vehicle speed, the vehicle posture and the safety sensor state, executing cutting track planning and collision checking under permission conditions, driving a cutting end executor to complete cotton plant terminal bud cutting, calling an image processing flow to recheck a cutting result after the operation is completed, generating an operation record packet containing time stamps, pose and track data, execution parameters, result labels and front-back comparison charts, and associating with pest risk reporting data and sending the pest risk reporting data to a central end when the conditions are met.
  5. 5. The cotton plant topping method based on image sensing and intelligent control according to claim 4, wherein the step S1 further comprises: s11, under unified triggering control, synchronously acquiring data by an industrial RGB camera, a vehicle speed sensor and an inertial measurement unit IMU, and binding the image after distortion correction with corresponding speed and posture information to form an original multisource synchronous data set; S12, calculating an intra-frame pose increment in the exposure time of the camera based on the vehicle speed and pose information, constructing an image plane homography matrix by adopting a local plane approximation model, and performing reverse mapping and resampling on an original image to realize image stabilization and shake correction; s13, estimating the motion blur direction and length according to the speed and posture information, constructing a motion blur check stabilized image to perform deblurring treatment, and implementing illumination normalization and vegetation color index enhancement to improve the separability of cotton plants and the background.
  6. 6. The cotton plant topping method based on image sensing and intelligent control according to claim 5, wherein the planar homography matrix in step S12 The calculation formula of (2) is as follows: , In the formula, As an internal reference matrix of the camera, For a fixed exogenous rotation of the IMU to the camera coordinate system, For relative rotation increments within the exposure time window, Is a camera external reference matrix Is to be used in the present invention, For relative translational increments within the exposure time window; The method is a transposition of a normal vector of a local plane unit under a camera coordinate system; Is a directed distance (scalar) to the local plane; Is the inverse of the camera reference matrix.
  7. 7. The cotton plant topping method based on image sensing and intelligent control according to claim 4, wherein the step S2 comprises the steps of: s21, inputting the enhanced image into a deep learning segmentation network, and performing pixel-level semantic segmentation on the top zone of the cotton plant to obtain a top zone semantic mask; S22, performing connected domain analysis and morphological filtering on the semantic mask to remove isolated noise points and non-target areas, obtaining a purification mask, constructing a top candidate band based on the upper edge boundary of the purification mask, calculating the minimum circumscribed rectangle of the top candidate band, generating a rectangular interest area according to a preset expansion coefficient, and performing boundary cutting, so that a top ROI data set is formed; s23, inputting the top ROI data set into a key point recognition network to obtain a candidate set of terminal bud key points, screening and uniquely selecting the candidate set according to a confidence coefficient threshold value and a maximum confidence coefficient criterion, and outputting a single terminal bud key point and the confidence coefficient thereof; S24, on the basis of the obtained terminal bud pixel coordinates, converting the key points from pixel coordinates to three-dimensional coordinates under a camera coordinate system by utilizing a camera internal parameter matrix and top depth or range parameters, and further utilizing external parameter of the camera to the machine body to perform coordinate transformation to obtain terminal bud space positions under the machine body coordinate system.
  8. 8. The cotton plant topping method based on image sensing and intelligent control according to claim 4, wherein the step S3 comprises the following steps: s31, performing super-pixel segmentation in an effective search domain to obtain a plurality of subareas, calculating color, texture and structural features of each subarea to form feature vectors, establishing an abnormality degree function based on background feature distribution of healthy blades, and sequencing and screening each subarea to obtain a candidate insect pest plaque set; s32, normalizing the clipping subgraphs corresponding to the candidate insect pest patches to a fixed scale and inputting the clipping subgraphs into an insect pest detection network to output insect pest categories, patch bounding boxes and detection confidence, and executing non-maximum suppression and consistency constraint on the multiple candidate results to obtain redundancy-removed detection results; And S33, extracting an anomaly characteristic in the detection target boundary box, carrying out weighted fusion on the anomaly and the detection confidence to obtain a pest risk score, and when the pest risk score exceeds a preset threshold value, executing mapping from pixels to ground plane coordinates on the basis of the camera internal and external parameters under the near ground plane assumption, and generating a risk report data packet comprising a time stamp, a pest category, a risk score, the boundary box, the ground coordinates and an evidence subgraph.
  9. 9. The cotton plant topping method based on image sensing and intelligent control according to claim 8, wherein step S33 is specifically to screen candidate pest patches in an effective search domain defined by the enhanced image and the canopy mask by using a super-pixel segmentation and mahalanobis distance abnormality function, and calculate a risk score by weighting and fusing abnormality and detection confidence after identifying by a pest detection network And generates and reports a risk data packet when the threshold is exceeded, The calculation formula of (2) is as follows: Triggering conditions: , In the formula, In order to have an a priori degree of anomaly, Is a priori anomaly.
  10. 10. The cotton plant topping method based on image sensing and intelligent control according to claim 4, wherein the step S4 specifically comprises the following steps: S41, generating a working window based on the space position and the confidence coefficient of the terminal bud and combining the vehicle speed, the attitude stability and the safety sensor state, and allowing to enter a topping execution flow when the confidence coefficient of the terminal bud is higher than a first threshold value and the safety condition is met, otherwise prohibiting execution and recording a reason code; S42, completing single shearing track planning and soft limit/collision detection in a working window, driving a shearing end effector to enter close alignment and execute shearing operation, automatically withdrawing to a safe position when abnormal conditions such as clamping stagnation or overtime occur, retrying according to limit times, and terminating the working and marking failure reasons when the retrying times exceed; And S43, after the shearing operation is finished, collecting the sheared images by using an industrial camera, rechecking the topping effect to form a job record packet comprising a timestamp, pose and track abstracts, an execution current or torque curve, a result label, ground coordinates and front-back contrast pictures, when the pest risk score reaches a second threshold value, associating the job record packet with corresponding pest risk data and reporting the job record packet to a central end, and when the communication bandwidth is limited, preferentially uploading the high-risk abstracts and locally caching detailed evidence data.

Description

Cotton plant topping system and method based on image sensing and intelligent control Technical Field The invention relates to the technical field of agricultural machine vision, in particular to a cotton plant topping system and method based on image sensing and intelligent control. Background Cotton is one of important economic crops in China, and the yield and quality of the cotton are directly related to raw material supply and economic benefits of farmers in the textile industry. In the growing process of cotton plants, topping operation is required in a proper period in order to promote nutrient distribution to fruit branches and improve the bearing rate of flowers and bolls. Topping can effectively inhibit top advantages and promote branch and leaf growth balance, and is a key link in cotton high-yield stable-yield management. However, the traditional manual topping method has the problems of high labor intensity, low efficiency, poor precision, strong manual dependency and the like, and is difficult to meet the development requirements of modern agricultural mechanization and intelligence. The existing automatic topping equipment mostly adopts a pure mechanical structure or adopts a single sensing means based on ultrasonic, infrared and the like to realize the detection and excision control of the terminal buds. Although the device improves the operation efficiency to a certain extent, the device has limited positioning precision and environmental adaptability, and especially under the conditions of complex field illumination, cotton plant posture change and interline difference, the problems of false detection, omission detection, topping deviation and the like are easy to occur. In addition, the lack of efficient image sensing and intelligent control strategies makes it difficult for the system to achieve accurate identification and dynamic decision control of the target terminal buds. For example, the invention patent with publication number CN118435797a discloses a full-automatic topping mechanism for cotton, and the system mainly comprises a movable frame, a rotary frame topping mechanism, a positioning mechanism and a terminal bud detection mechanism. According to the scheme, the multi-point topping operation is realized through the rotating disc, the radius adjusting assembly and the lifting topping assembly, different cotton plant spacing and height differences are adjusted through a mechanical structure, and automatic topping to a certain extent can be completed. However, the mechanism mainly relies on mechanical sensing and cylinder control to realize height fixing and roof picking, lacks real-time visual identification and intelligent decision making capability, has a complex overall structure and high degree of coupling between modules, and does not have self-adaptive capability on different illumination conditions and cotton plant posture changes. Therefore, a cotton plant topping system and method based on image sensing and intelligent control are urgently needed, and an efficient and reliable intelligent solution is provided for cotton field management. Disclosure of Invention The invention aims to provide a cotton plant topping system and a cotton plant topping detection method based on image sensing and intelligent control, which are used for realizing automatic identification and positioning of cotton plant terminal buds by integrating an industrial camera, an Inertial Measurement Unit (IMU) and a vehicle speed sensor, and realizing automatic detection, intelligent decision and accurate execution of cotton plant topping by executing image processing and path planning by an industrial control computer and driving a lifting and shearing mechanism to finish accurate topping. The cotton plant topping system comprises a machine body, an image and gesture sensing module, a topping executing module and an image processing and intelligent control module, wherein the machine body comprises a moving platform, a metal frame and travelling wheels, the moving platform, the metal frame and the travelling wheels are used for bearing all functional modules and moving in cotton fields, the image and gesture sensing module comprises an industrial RGB camera, an inertial measurement unit IMU and a vehicle speed sensor, the industrial RGB camera, the inertial measurement unit IMU and the vehicle speed sensor are arranged above and in front of a machine body shell and are used for synchronously acquiring cotton plant canopy images and vehicle motion gesture information, the topping executing module comprises a lifting module, a guide sliding rod, a shearing driving assembly and a shearing end effector, the lifting module and the guide sliding rod are used for driving and restraining the shearing end effector to move in the vertical direction, and the shearing end effector is used for shearing cotton plant top buds, and the image processing and intelligent control module comprises an industrial control