研究目的：立定跳远运动中，学生的动作姿态对其远度具有一定的影响，当前教学内容及教学计划的制定，主要依靠体育教师人力观察及个人经验，存在针对性较差、主观性较强等问题。借助计算机视觉技术辅助体育教师追踪、识别、分析学生动作姿态，从而帮助教师进行更加客观、科学的决策，改善学生立定跳远动作姿态。本研究基于人体姿态识别技术开发立定跳远动作诊断与评价系统，旨在建立一套立定跳远错误动作智能识别系统，此系统能够针对学生立定跳远的错误动作进行精准识别与评价，帮助教师更高效、客观、准确的提出针对性的训练方案。

研究方法：本研究采用系统开发法、实验法等设计开发了一套基于人体姿态识别技术的立定跳远动作诊断与评价系统。系统开发法，首先采集大量立定跳远动作视频，并将其进行错误动作分类，将错误动作导入系统，系统进行深度学习，建立立定跳远错误动作大模型。通过系统开发法，识别诊断学生立定跳远错误动作类型。实验法，选取德州市第九中学九年级（25）班24名学生作为研究对象，分为实验组（n=12）与对照组（n=12），实验组进行为期4周的实验干预，每周训练3次，每次15分钟。通过实验法验证立定跳远动作诊断与评价系统应用在体育教学中的可行性及其教学效果。

研究结果：本研究（1）通过系统置信度测试，系统可以准确反馈学生的错误动作类型，并在识别过程中提供身体各关节角度、重心移动轨迹等运动学指标。

（2）选取肩关节角度、髋关节角度、膝关节角度等运动学指标，经过立定跳远动作诊断与评价系统对实验组的干预，实验组和对照组均表现出显著性差异。起跳前最后一次预摆阶段中，实验组和对照组左右膝关节、左右髋关节角度平均值及标准差分别为81.96±9.37、101.48±20.27（P<0.05），84.76±10.67、101.30±21.09（P<0.05），49.82±9.66、67.15±5.57（P<0.05）,50.16±9.60、65.72±5.00（P<0.05）；起跳时髋膝踝伸展阶段中，左右膝关节、左右肩关节角度分别为164.67±7.45、151.67±10.24（P<0.05），165.38±7.26、151.36±14.21（P<0.05），32.30±17.60、55.80±17.50（P<0.05），33.14±18.04、55.61±16.72（P<0.05）；腾空时收腹团身及举腿阶段中，左右膝关节、左右髋关节角度分别为147.40±5.56、135.13±16.93（P<0.05），149.59±8.33、134.42±16.98（P<0.05），51.75±4.65、68.92±15.77（P<0.05），49.50±4.33、49.50±4.33（P<0.05），以上各关节角度差异性具有统计学意义，说明立定跳远动作诊断与评价系统对提高动作的规范性具一定的帮助。此外，研究结果也表明，预摆时肩关节角度、起跳时髋关节角度并无显著差异，可能学生对这两种动作掌握较快。

研究结论：（1）本研究开发了基于人体姿态识别技术的立定跳远动作诊断与评价系统，系统可以自动识别立定跳远动作的错误类型。

（2）本研究证明了该系统可以作为立定跳远教学与评价的辅助工具。

（3）立定跳远动作诊断与评价系统在体育教学中应用，构建出一种人机协同体育教学新模式，提高了教师及学生的感知能力、认知能力以及决策能力，增强了体育教学过程中的直观性、科学性、高效性。

Research purpose:In the standing long jump, students' movement and posture have a certain influence on their distance. The current teaching content and teaching plan mainly rely on the human observation and personal experience of physical education teachers, and there are problems such as poor pertinence and strong subjectivity.With the help of computer vision technology, PE teachers help to track, identify and analyze students 'movements, so as to help teachers to make more objective and scientific decision-making and improve students' standing long jump posture. This study is based on human body posture recognition technology development of standing long jump motion diagnosis and evaluation system, aims to establish a set of standing long jump error intelligent recognition system, this system for students standing long jump error accurate identification and evaluation, help teachers more efficient, objective and accurate targeted training plan.

Research method: A set of standing long jump motion diagnosis and evaluation system based on human posture recognition technology was developed by using system development method and experimental method. In the system development method, a large number of standing long jump action videos are collected, and the wrong movements are classified, the wrong movements are imported into the system, the system conducts deep learning, and a large model of standing long jump error action is established. Through the system development method, identify and diagnose the students' standing long jump error action type. In experimental method, 24 students from Grade 9 (25) of Dezhou No.9 Middle School were selected as research objects and divided into experimental group (n=12) and control group (n=12). The experimental group conducted experimental intervention for 4 weeks, training 3 times a week for 15 minutes. The feasibility and teaching effect of the diagnosis and evaluation system in physical education teaching are verified.

Results: This study (1) Through the system confidence test, the system can accurately feedback the wrong action types of students, and provide kinematic indicators such as the Angle of the joint and the movement track of the center of gravity of the body during the identification process.

(2)Selected kinematic indicators such as shoulder Angle, hip Angle and knee Angle, and selected the experimental group through the intervention of the standing long jump movement diagnosis and evaluation system, both the experimental group and the control group showed significant differences. In the last pre-swing stage before takeoff, The mean and standard deviation of the left and right knee and hip angles in the experimental and control groups were 81.96 ± 9.37 and 101.48 ± 20.27, respectively (P <0.05), 84.76±10.67,101.30±21.09(P<0.05), 49.82±9.66,67.15±5.57(P<0.05), 50.16±9.60,65.72±5.00(P<0.05); During the hip, knee and ankle extension stage, The angles of left and right knee and left and right shoulder were 164.67 ± 7.45 and 151.67 ± 10.24 respectively (P <0.05), 165.38±7.26,151.36±14.21(P<0.05), 32.30±17.60,55.80±17.50(P<0.05), 33.14±18.04,55.61±16.72(P<0.05); During the stage of abdominal body gathering and leg lifting, Angangles of left and right knee and left and right hip were 147.40 ± 5.56 and 135.13 ± 16.93 respectively (P <0.05), 149.59±8.33,134.42±16.98(P<0.05), 51.75±4.65,68.92±15.77(P<0.05), 49.50±4.33,49.50±4.33(P<0.05), In addition, the study results also show that there is no significant difference between the shoulder Angle and the hip Angle during the pre-swing, and the students may master these two movements quickly.

Conclusion: (1) This study has developed the diagnosis and evaluation system of standing long jump movement based on human posture recognition technology, which can automatically identify the wrong types of standing long jump movement.

(2) This study proves that the system can be used as an auxiliary tool for the teaching and evaluation of the standing long jump.

(3) The standing long jump movement diagnosis and evaluation system is applied in physical education teaching to construct a new mode of man-machine collaborative physical education teaching, which improves the perception ability, cognitive ability and decision-making ability of teachers and students, and enhances the intuitive, scientific and efficient in the process of physical education teaching.

KEY WORDS:Posture Recognition,Standing Long Jump,Motion Diagnosis,Teaching evaluation