In complex scenes, illumination variation and changes in the target′s scale may lead to loss of targets or mis-tracking.To solve the problem, this paper proposes a structured multi-target tracking algorithm adaptive to scale and illumination changes.Multi-scale Retinex algorithm is used to preprocess the image sequence.Structure Preserving Object Tracking (SPOT) algorithm is used to track multiple targets, so as to determine the optimal position of each target in the new frame.Discriminant Scale Space Tracking (DSST) algorithm is used to train the scale filter, whose maximum value is used to determine the optimal scale of each target in the new frame by taking the optimal position of each target in the new frame as the center.Stochastic Gradient Descent (SGD) and bilinear interpolation are used to update the weight of the feature classifier.The experimental results show that the proposed multi-target tracking algorithm has good robustness and accuracy in dealing with illumination variation and changes in the target′s scale.