In order to obtain the distribution information of the kiwifruit orchards in high spatial resolution remote imagery fast and accurately, a hybrid method for automatic detection of kiwifruit orchard based on wavelet transform and random forest classification algorithm was proposed. Firstly, a wavelet transform based texture extracting process was carried out on the QuickBird panchromatic band by means of a two level decomposition with coif5 biorthogonal wavelet function, and the multi-scale wavelet textures were further derived from the energy characteristics of the wavelet coefficients in each sub-band. Secondly, the wavelet textures and spectral features were combined to construct the classification feature vectors. Finally, the kiwifruit orchard distributions were automatically delineated through land cover classification by using the random forest ensemble technique. The wavelet textures were found to be more effective in identifying kiwifruit orchard compared with the multi spectral features, gray level co-occurrence matrix (GLCM) textural features and fractal textural features. There was an obvious increase in kiwifruit orchard extracting accuracy (Fk) and overall classification accuracy (OA) when spectral features were combined with textural features compared with spectral-only and texture-only features. The highest classification accuracies were achieved by the integration of spectral features and the multi-scale wavelet texture features (spectral + wavelet TF) with Fk of 95.30% and OA of 94.46%, which was 6.70% and 2.88% higher respectively than those of the results of spectral+ GLCM features and 13.43% and 6.98% higher respectively than those of spectral + fractal features. Among the three classifiers used, the random forest classifier demonstrated the best performance in terms of OA and Fk, followed by support vector machine classifier and the maximum likelihood classifier under the same features. The extracted area of kiwifruit orchard was also assessed by the visual interpretation results and the relative error was less than 7%. An apple orchard extracting experiment in another test region was carried out by using the same method, and the results indicated that the method had good applicability.