In recent years, the application of wearable inertial sensors in sports has marked a significant advancement in the analysis of athletes' physical performance. Due to their small size, these sensors are considered favorable for maintaining freedom of movement, thus providing a good balance between practicality and measurement accuracy. This thesis focuses on the evaluation and calibration of the inertial sensor (Inertial Measurement Unit - IMU) K-AI, developed by K-Sport (Fano, Italy), for the analysis and measurement of jump height in sports. The sensor was positioned on the athlete’s back, inside a specially designed vest to enhance performance during jump sessions. To validate the data collected by the sensor, they were compared with data acquired from a Nikon D7200 camera (Nikon Corporation, Tokyo, Japan), used as the reference device. The subjects involved in this study were 31 professional basketball and volleyball athletes, who performed different types of jumps (squat jump, block jump, jump shot) at distances of 4, 5, and 6 meters from the camera. Additionally, at a distance of 5 meters, jumps were performed at three varying intensities (high, medium, low). The camera was properly calibrated to establish a pixel/cm conversion factor, which was essential for the analysis. Moreover, a static calibration was performed on three stationary subjects at distances ranging from 4 to 9 meters by measuring the actual distances between anatomical landmarks and comparing them with values calculated using the MediaPipe Pose Landmarker library in Python. Video analysis was conducted by identifying pelvic landmarks using Python's MediaPipe library. From these, the midpoint was calculated and used to identify the peak values of the signal. Jump height was estimated as the difference between the peak values of the signal obtained from the tracking of the pelvic midpoint and the threshold value corresponding to the subject's standing pelvic position. For the analysis of jump height using the IMU sensor, flight time was derived by segmenting the signal into windows isolating a single jump (identifying negative peak values) and evaluating the intersection of the curve with the zero-acceleration threshold. Using statistical tools such as residual distribution, the Bland-Altman plot, and correlation plot, the results were obtained: from the histogram, a mean of -0.12 m and a standard deviation of 0.04 m were observed, while the 95% confidence interval and the Pearson correlation coefficient were [-0.20; -0.03] m and 0.95, respectively. Finally, through regression analysis, the calibration curve of the K-AI IMU device was obtained, thus completing the study.
Negli ultimi anni, l’applicazione di sensori inerziali indossabili in ambito sportivo ha segnato un punto di sviluppo per le analisi delle performance fisiche degli atleti. Per via delle loro dimensioni ridotte, sono considerati favorevoli alla libertà di movimento, garantendo così il giusto compromesso tra praticità e accuratezza delle misure. La presente tesi si concentra sulla valutazione e calibrazione del sensore inerziale (Inertial Measurment Unit - IMU) K-AI sviluppato da K-Sport (Fano, Italia), per l’analisi e la misura dell’altezza del salto in ambito sportivo. Il sensore è stato posizionato sulla schiena degli atleti, all’interno di una pettorina progettata appositamente usando per favorire le prestazioni durante le sessioni di salto. Per validare i dati raccolti dal sensore, essi sono stati confrontati con i dati acquisiti dalla fotocamera Nikon D7200 (Nikon Corporation, Tokyo, Giappone), che rappresenta il dispositivo di riferimento. I soggetti coinvolti in questo studio sono 31 atleti professionisti di pallacanestro e pallavolo, i quali hanno eseguito diverse tipologie di salti (squat jump, salto a muro, salto tiro a canestro) a distanze di 4, 5 e 6 m dalla fotocamera. In aggiunta, ad una distanza di 5 m, sono stati effettuati salti a tre intensità variabili (alta, media, bassa). La fotocamera è stata adeguatamente calibrata, per poter stabilire un fattore di conversione pixel/cm, fondamentale nelle analisi. Inoltre, è stata condotta una taratura statica su tre soggetti immobili, a distanze comprese tra 4 e 9 m, misurando le distanze reali tra i punti di repere dei soggetti e confrontandole con i valori calcolati usando la libreria MediaPipe Pose Landmarker di Python. L’analisi video è stata svolta individuando i punti di riferimento del bacino usando la libreria MediaPipe di Python e, a partire da questi si è calcolato il punto medio, usato successivamente per individuare i picchi massimi del segnale. L’altezza del salto è stata stimata come differenza tra i picchi massimi del segnale ottenuto dal tracciamento del punto medio del bacino e il valore di soglia corrispondente alla posizione del bacino del soggetto immobile. Per l’analisi dell’altezza del salto tramite sensore IMU, invece, viene ricavato il tempo di volo andando a segmentare il segnale in finestre che isolano un singolo salto (individuando i picchi massimi negativi) e andando a valutare l’intersezione della curva con la soglia ad accelerazione nulla. Tramite strumenti statistici come la distribuzione dei residui, il grafico di Bland-Altman e quello di correlazione sono stati ottenuti i risultati: dall’istogramma si osserva una media di -0,12 m e una deviazione standard di 0,04 m, mentre l’intervallo di confidenza al 95% ed il coefficiente di correlazione di Pearson valgono rispettivamente [-0,20; -0,03] m e 0,95. Infine, tramite l’analisi di regressione è stato possibile ricavare la curva di calibrazione del dispositivo IMU K-AI, portando a termine lo studio.
CARATTERIZZAZIONE METROLOGICA DI UN DISPOSITIVO INDOSSABILE PER MISURE DI SALTI IN AMBITO SPORTIVO: MISURE IN CONDIZIONI OPERATIVE
BONTEMPO, SERENA
2024/2025
Abstract
In recent years, the application of wearable inertial sensors in sports has marked a significant advancement in the analysis of athletes' physical performance. Due to their small size, these sensors are considered favorable for maintaining freedom of movement, thus providing a good balance between practicality and measurement accuracy. This thesis focuses on the evaluation and calibration of the inertial sensor (Inertial Measurement Unit - IMU) K-AI, developed by K-Sport (Fano, Italy), for the analysis and measurement of jump height in sports. The sensor was positioned on the athlete’s back, inside a specially designed vest to enhance performance during jump sessions. To validate the data collected by the sensor, they were compared with data acquired from a Nikon D7200 camera (Nikon Corporation, Tokyo, Japan), used as the reference device. The subjects involved in this study were 31 professional basketball and volleyball athletes, who performed different types of jumps (squat jump, block jump, jump shot) at distances of 4, 5, and 6 meters from the camera. Additionally, at a distance of 5 meters, jumps were performed at three varying intensities (high, medium, low). The camera was properly calibrated to establish a pixel/cm conversion factor, which was essential for the analysis. Moreover, a static calibration was performed on three stationary subjects at distances ranging from 4 to 9 meters by measuring the actual distances between anatomical landmarks and comparing them with values calculated using the MediaPipe Pose Landmarker library in Python. Video analysis was conducted by identifying pelvic landmarks using Python's MediaPipe library. From these, the midpoint was calculated and used to identify the peak values of the signal. Jump height was estimated as the difference between the peak values of the signal obtained from the tracking of the pelvic midpoint and the threshold value corresponding to the subject's standing pelvic position. For the analysis of jump height using the IMU sensor, flight time was derived by segmenting the signal into windows isolating a single jump (identifying negative peak values) and evaluating the intersection of the curve with the zero-acceleration threshold. Using statistical tools such as residual distribution, the Bland-Altman plot, and correlation plot, the results were obtained: from the histogram, a mean of -0.12 m and a standard deviation of 0.04 m were observed, while the 95% confidence interval and the Pearson correlation coefficient were [-0.20; -0.03] m and 0.95, respectively. Finally, through regression analysis, the calibration curve of the K-AI IMU device was obtained, thus completing the study.File | Dimensione | Formato | |
---|---|---|---|
Tesi_SB.pdf
embargo fino al 18/01/2027
Descrizione: Tesi con frontespizio non firmato
Dimensione
2 MB
Formato
Adobe PDF
|
2 MB | Adobe PDF |
I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.12075/22080