Intelligent monitoring technology for the water flow of oil drilling can realize the automatic monitoring of gaseous pollutant from oil drilling and minimize the cost of manual monitoring to the greatest extent. The existing feature extraction methods cannot describe the change process of water flow, it is difficult to obtain abnormal samples and fully enumerate them, and the fusion layer information is not fully utilized. In order to solve the problems, a new water flow abnormal data detection algorithm was proposed. Firstly, a new feature representation method named shape flow was proposed. Then, the classic anomaly detection unsupervised neural network GANomaly was optimized into a residual structure. Finally, a feature fusion layer was added to the GANomaly to improve the learning ability of neural network. Experimental results show that, the detection accuracy of the improved algorithm reaches 95%, which is 5 percentage points higher than that of the GANomaly algorithm. The proposed algorithm can be applied to the detection of abnormal water flow data in different scenarios, and can overcome the influence of fog on the experimental results.

Concerning the problem that pedestrians would be partially or seriously shaded by each other in video monitoring, this paper proposed a people counting algorithm based on human body skeleton feature. At first, the initial human skeleton was extracted by morphological skeleton extraction algorithm. Then the optimal skeleton feature was obtained by eliminating outliers and pseudo branches. Finally, this paper established a head detection response rule through analyzing the characteristics of skeleton in head areas to detect the head of pedestrian, and completed people counting by counting the heads of pedestrians. The experimental results show that the algorithm can solve the problems of partial and serious shading in video monitoring. For relatively sparse scene, the overall people counting accuracy rate of the algorithm is about 95%.

In view of the correction error problem which is caused by some factors such as dithering, the authors presented a new optical tilt correction method based on grating projection. The method was based on the analysis of each pixel of the data array in a sequence of fringe patterns having multiple frequencies, and setup model for pixel coordinates and pixel-slope. Then skew angles of fringes were calculated by trigonometry with the relationship between tilt angle and pixel-slope. At last, tilt correction was realized. The experimental results show that, the algorithm is capable of accurately detecting angle within the range [-90°,90°],accuracy is 99%. Compared with other algorithms such as Hough transform, the proposed algorithm improves precision and accuracy significantly.

Behavior analysis based on vision is one of the important research topics in image processing, pattern recognition,etc, and it has wide application prospects on public security and military field. For the problems of a fixed camera such as lack of the single feature description, motion occlusions, holes and shadows, the paper proposed a behavior recognition algorithm which combines space-time topological feature with sparse expression. It used random projection to get a space-time topological feature of strong cohesion, high distinction and low dimension, which fused topology structure, geometric invariant and space-time Poisson information. The noise-adding sparse mechanism resolving problems by simulating human was combined to identify behaviors of human body in a close-range monitor scene. The experimental results show that the recognition rate of space-time topological feature is 12.79% higher than that of single one. The recognition rate of this proposed algorithm is only 6.15% down in a noisy scene, and that for multi-behavior reaches 87.78%. This algorithm has the properties of strong description for space-time feature, higher robustness against noise and high efficiency for behavior recognition.