CN-116385479-B - Animal three-dimensional moving track tracking measurement recording method based on depth camera

Abstract

The invention discloses a three-dimensional animal movement track tracking measurement recording method based on a depth camera, which mainly comprises the steps of S1, mounting the depth camera at the front side of a transparent cage for 1.5-2 meters, adjusting a proper distance according to the size of the transparent cage, enabling a visual field to cover the cage as large as possible, S2, accurately positioning eight vertexes of the inner space of the cage, S3, acquiring and removing background noise points in real time, S4, synchronously recording four-compartment images by calling an infrared camera of Kinect2.0 while measuring and tracking three-dimensional track data in real time, and being suitable for measuring relevant behavioral index measurements such as accurate movement quantity, frequency, position preference, movement track, whether the measurement of the indexes is beneficial to modeling and evaluation of various senile degenerative diseases, drug effects and rehabilitation therapies.

Inventors

• HE JING
• WEI JINGKUAN

Assignees

• 昆明云岸数字科技有限公司

Dates

Publication Date

20260508

Application Date

20220816

Claims (2)

1. 1. A method for accurately tracking, measuring and recording three-dimensional moving tracks of a single experimental animal based on a depth camera is characterized by comprising the following steps: S1, mounting a depth camera at 1.5-2 meters in front of the side of a transparent cage, and adjusting a proper distance according to the size of the cage, so that the picture is as large as possible on the premise that the field of view covers the cage; s2, accurately positioning eight vertexes of the inner space of the cage; s3, obtaining and eliminating a background noise point in real time; S4, synchronously recording the four-grid picture by calling an infrared camera of Kinect2.0 while measuring, tracking and recording three-dimensional track data in real time, so that an experimenter can conveniently and comprehensively review the data; the vertex positioning method of the S2 specifically comprises the following steps: S2.1, displaying real-time data of three-dimensional point clouds of a camera on a three-dimensional picture, and ensuring that an operator can obtain visual picture feedback when eight vertexes of the cage are adjusted and positioned; s2.2, calculating and feeding back the point cloud pixel points inside and outside the calibration space in real time according to the space positions of the eight vertexes; s2.3, utilizing reflective cloth to assist the positioning operation of far-end vertexes and vertexes with other sight-impaired, wherein the accurate positions of three-dimensional point clouds of vertexes in a cage space are often not very clear, and the reflective cloth can be used for assisting in calibration positioning by virtue of the characteristics of high brightness in an infrared image and no point cloud data, is placed at corners of the vertexes and is matched with three-dimensional dragging rotation of a mouse; S2.4, feeding back the calibrated physical positions of the vertexes and the corresponding physical distance values among the vertexes in a three-dimensional picture to an operator in real time on the picture, and further setting the accurate vertex positions of the cages by comparing the actual sizes of the cages; the method for calibrating the point cloud pixel points inside and outside the space comprises the following steps: s2.2.1 firstly, calculating the corresponding minimum/maximum depth values of the front/rear two surfaces according to the X/Y position of any point and the space linear mapping relation; S2.2.2, calculating the minimum/maximum horizontal axis position on the left/right two surfaces corresponding to the point and the minimum/maximum height position on the upper/lower two surfaces corresponding to the point according to the proportion position of the depth value Z of the point on the minimum/maximum depth value according to the space linear mapping relation; S2.2.3 finally judging whether the X/Y/Z three-dimensional positions of the points are all in the corresponding minimum/maximum range, if yes, marking the points as yellow in real time, and if no, marking the points as blue; The specific method for acquiring and eliminating the background noise point in the S3 in real time comprises the following steps: S3.1, generating a three-dimensional data array with a proper size according to the recording space framed by the fixed-point positioning method step, and recording and storing noise data generated by an inherent object in the recording space; S3.2, in the process of tracking and recording, for each point cloud data in a recording space, traversing 3 x 3 positions near the data point in a three-dimensional noise data frame, and judging the data point as one of current effective data values if no noise point in the recording is found; The method for optimizing the running efficiency comprises the steps of limiting the range, spacing and sampling and calculating the rotation, and specifically comprises the steps of firstly framing an effective range by using the XYZ range values of the appointed eight vertexes, secondly sampling by spacing n points each time, and adding an auxiliary rotation parameter which changes from 0 to n one by one every frame so as to ensure that all pixel points of every n frames are calculated once.
2. 2. The method for accurately tracking, measuring and recording the three-dimensional moving track of the single experimental animal based on the depth camera is characterized in that the four-grid pictures are respectively 1) infrared image video data, 2) cage point positions on a three-dimensional canvas and tracking point position display, 3) infrared image video data and cage point positions on the three-dimensional canvas and tracking point position display, 4) experiment numbers, dates, recorded real-time elapsed time and real-time three-dimensional track values.

Description

Animal three-dimensional moving track tracking measurement recording method based on depth camera Technical Field The invention belongs to the technical field of detection and tracking of moving objects, and particularly relates to a method for accurately tracking, measuring and recording a three-dimensional moving track of a single experimental animal based on a depth camera. Background Under the background of social aging, more and more biomedical experiments need to measure the activity change of model-building experimental animals of various degenerative diseases and the accurate three-dimensional activity track and other behavioral data to make evaluation indexes of the experiments of disease model building, rehabilitation and the like. The existing method for automatically analyzing the three-dimensional track of the experimental animal mainly depends on the two-dimensional picture of a common camera, and the spatial position of the animal in the cage is roughly estimated by identifying the two-dimensional picture based on a depth model. The optimized scheme comprises the steps of adopting a plurality of common cameras to shoot the measured object at the same time from different angles, and estimating the position of the target object in the three-dimensional space based on depth model identification in principle as a single camera. The existing method for tracking the three-dimensional track of the experimental animal in the cage space based on the depth camera is not available. SDK2.0 of Microsoft Kinect 2.0 provides the nearest three-dimensional point cloud information within the field of view captured by the depth camera and within the effective depth range. However, the practical application of the data to the measurement and recording of the three-dimensional trajectories of experimental animals in cages faces several key problems including 1, how to accurately position and restore the positions of eight vertices of the internal space of the cage in the recorded images (three dimensions) to accurately frame the recorded space, 2, how to remove the images of the objects inherent in the measured space on the tracked measurement data to make the accuracy of the data within + -1 cm, 3, how to improve the acquisition and operation efficiency, ensure that accurate data is tracked in real time, synchronously record the tracked data and the image data, and maintain a frame rate of more than 30 frames per second. Disclosure of Invention In order to solve the technical problems, the invention provides a depth camera-based method for accurately tracking, measuring and recording three-dimensional activity tracks of single experimental animals, which is suitable for measuring relevant behavioral index measurements such as activity amount/movement/muscle tension of the experimental animals with three-dimensional activity in a cage space, such as accurate activity amount, bouncing frequency, position preference, movement tracks, whether activity patterns exist, track abnormality and the like, and the measurement of the index is helpful for modeling various senile degenerative diseases, medication effects, and effect evaluation of rehabilitation. In order to achieve the technical purpose, the invention is realized by the following technical scheme: A method for accurately tracking, measuring and recording three-dimensional moving tracks of a single experimental animal based on a depth camera comprises the following steps: S1, mounting a depth camera at 1.5-2 meters in front of the side of a transparent cage, and adjusting a proper distance according to the size of the cage, so that the picture is as large as possible on the premise that the field of view covers the cage; s2, accurately positioning eight vertexes of the inner space of the cage; s3, obtaining and eliminating a background noise point in real time; s4, synchronously recording the four-grid picture by calling an infrared camera of Kinect2.0 while measuring, tracking and recording three-dimensional track data in real time, so that an experimenter can conveniently and comprehensively review the data; preferably, the vertex positioning method of S2 specifically includes: S2.1, displaying real-time data of three-dimensional point clouds of a camera on a three-dimensional picture, and ensuring that an operator can obtain visual picture feedback when eight vertexes of the cage are adjusted and positioned; s2.2, calculating and feeding back the point cloud pixel points inside and outside the calibration space in real time according to the space positions of the eight vertexes; s2.3, utilizing reflective cloth to assist the positioning operation of far-end vertexes and vertexes with other sight-impaired, wherein the accurate positions of three-dimensional point clouds of vertexes in a cage space are often not very clear, and the reflective cloth can be used for assisting in calibration positioning by virtue of the characteristics of high brightness in an in