Binary Histogram of Oriented Gradients Based Dynamometer Card Recognition

Surface dynamometer card of pumping unit is a closed curve of load and displacement values in a pumping cycle, which reflects the operating condition of the sucker-rod pumping unit. Automatic analyzing and diagnosing the dynamometer card plays an important role in the process of oil and gas production. This paper proposes a robust Binary Histogram of Oriented Gradients (BHOG) feature for dynamometer card recognition. It is much simpler and more effective than the general HOG feature and has great performance. BHOG feature represents the binary image by 4 kinds of texture (corresponding to 4 bins of histogram in HOG), could highly reduce the feature size. Besides, the process of feature extraction has also been simplified. Experiment shows that the proposed BHOG feature is 5.32 times faster than the general HOG feature, and achieves 98.19% accuracy in the dynamometer card dataset.