Abstract: In the face of strong electromagnetic interference, industrial environment or military electronic countermeasure defense and other complex environment, sensor measurement is easily affected by impulse noise. This reduces the performance of cardinality balanced multi-object multi-Bernoulli (CBMeMBer) filters. In order to reduce the impact of impulse noise on the filter, a cardinality balanced multi-object multi-Bernoulli based on impulse noise fusion-CBMeMBer(INF-CBMeMBer) is proposed. The measured noise is modeled as a Gaussian mixture model, which is characterized by the weighted sum of Gaussian distributions of stationary Gaussian noise with small covariance and impulse noise with large covariance. The expectation-maximization (EM) algorithm is used to dynamically calculate the parameters of noise, including mean, covariance and impulse noise occurrence probability. The multi-target states of stationary Gaussian noise and pulse noise are updated respectively in the updating step of multi-Bernoulli, and the weights of stationary Gaussian noise and impulse noise are calculated by the likelihood function of Gaussian distribution, and the multi-target states are fused by weights. The simulation results verify the effectiveness of the proposed filter, and the filter has a good tracking effect in the scene affected by impulse noise.