The extraction of forest texture feature from color forest aerial remote sensing images is an important part of the automatic interpretation of forest types achieved by computer. In this dissertation, five kinds of typical forests sample images, including natural forest, plantation forest, young-aged forest, sutiable for forest land and the combination of forest and farm crop, are selected from the color aerial remote sensing images of Beijing area with ground resolution of 0.8 meters. By understanding the theory, method and technology of image texture feature extraction having been existed, Gray co-occurrence matrix, Local binary pattern and 2D Gabor filters are chosen.. The corresponding algorithm and the texture feature parameters are designed to distinguish this several kinds different forest texture. Then the validity of various methods is analyzed. The main works of the thesis is as follows:1. The extraction of forest texture feature applying the co-occurrence matrix method is researched. The impact of pixel distance over characteristic value is analyzed. Three-dimensional vector composed with energy, contrast and entropy are built to describe the characteristics of forest texture. Based on the relevance of color component of RGB color image, the colored correlation matrix are introduced to the texture extraction of color image. Two values applied to recognize the generic difference and inner-consistency are proposed to evaluate the efficient of distinguishing forest types using texture features in different component color matrix.2. The extraction of forest texture feature applying Local binary pattern is researched. Several improvements, including Local binary pattern operator combined with local variance, Local binary pattern operator combined with Local contrast and Uniform pattern, are designed and carried out. For the color images, the local contrast is melted into the opponent color Local binary pattern to form a new texture parameter, which can distinguish the different forest texture more effectively.3. The extraction of forest texture feature applying Two-dimensional Gabor filter method is researched.. According to the nature of 2D Gabor filter very similar with human vision nerve cell inthe frequency responses range and the frequency response bandwidth range, we proposed two design principle for our 2D Gabor filter. One is that the response existence of two filters, which have same phase and neighboring center frequency, should has enough overlap of frequency along radial direction. The other is that two neighboring Gabor filter, which have same center frequency and neighboring phase, should have enough overlap of phase. After analysis to 5 kind of forests types frequency spectrum the definite filter direction is determined. A group of Gabor filters to extract different types of forest image texture features is designed. Three texture parameters, including the amplitude of energy, cross-color feature and invariable feature, are extracted form the filtered images. The distinguish ability of Cross-color features is most strongest among of them.4. A non-parametric density estimates method in LUV color space to color cluster, which do not need any prior knowledge, is researched. The segmentation of the forest region in the image is carried out efficiently. At final, a theory frame, which integrates forest texture feature with the color feature to enhance accuracy of regional segmentation and identification of the forest types, is proposed.