Shupe effect has a significant influence on a precision of the fiber-optic gyro (FOG). The effect itself is an emergence of a parasite signal at the gyro output when external temperature changes or during device self-heating. In order to decrease influence of the effect mentioned (parasite phase shift), one uses special winding types combined with the potting compound. As a result, FOG precision significantly depends on the winding quality of its fiber coil. Ideally, the wound coil should have no winding defects and the fiber should form a periodic uniform structure. When winding FOG coils using a winding machine, it’s advisable to use automated quality control based on the machine vision. In order to use machine vision one has to provide high quality of image form camera in the first place. The goal of the study is a development of FOG coil winding defect detection technology and estimation of the efficiency of different image filtering algorithms for real-time processing of the machine vision camera’s signal. In addition to the traditional Gaussian filter, wavelet filtering is proposed for video frames de-noising. As a base of real-time defect detection algorithm, detection and analysis of the winding profile during winding process is used. Results of the nature modeling of the winding process of a polymer fiber onto a coil carcass revealed that the algorithm proposed allows one to detect coil winding defects with a high probability, and that it can be recommended for use during manufacturing of fiber-optic gyro coils.