Soft tactile sensors have the ability to infer more physical properties of an object relative to classical optical motion-capture systems. Three marker densities in a tactile sensor array (the Motion Capture Pillow, MCP) were evaluated for tracking two rotary motions using a weighted mannequin head. The Kanade–Lucas–Tomasi algorithm was employed to track head movements using three silicone sheets, each embedded with different marker spacings (5, 10, and 15 mm). The averaged Spearman’s correlation slightly changed from 0.80 (for 10 mm spacing) to 0.67 (for 5 mm spacing) for pitch motion and from 0.68 (for 10 mm spacing) to 0.59 (for 5 mm spacing) for roll motion of the mannequin head with respect to the MCP’s frame. A correlation of +1.0 being the strongest positive correlation and 0.0 being weak correlation. The MAE reduced by 12.9% from matrix with 10 mm spacing to 5 mm spacing for pitch motion, and by 2.9% for roll motion. This established a foundation for further tuning the sensor using a higher density of the sensing matrix. The relatively sparsely dense sensor matrix with 15 mm spacing had minimal impact on the tracking performance of the sensor. Sources of noise were narrowed down to hysteresis, and boundary conditions. These results demonstrated the influence of marker density on the object tracking abilities of an optical soft tactile sensor, and established a basis for future optimisation.