Skyline Detection Algorithm Based On Gray Entropy

Skyline detection plays an important role in many aspects,but the current detection algorithm can only satisfy the skyline detection in a simple background.In the general outdoor smoke detection system,due to the camera's jitter,atmospheric flow and other factors,the cloud of the detection algorithm near the skyline is falsely reported as smoke,resulting in a decrease in the accuracy of the smoke alarm.Skyline detection and preprocessing are important means to reduce false positives.In information theory,signals are classified into two types: deterministic and random.Image signals are classified as random signals due to background complexity and noise interference uncertainty,and entropy is the only method that can determine the amount of random signal changes.In view of this,this thesis proposes a skyline detection algorithm based on gray entropy in images.The method firstly performs pre-treatment of the obtained image by defogging,and uses Dr.He Kaiming's dark channel defogging method.,and then the image of the preprocessed image is analyzed.After the a priori of entropy and Gaussian filtering is used for denoising,the gray entropy of each pixel in the direction of the image column is calculated.The calculation method is as follows: the point is divided into two points,the column is divided into upper and lower parts,and the sum of the gray entropies of the upper and lower parts is subtracted by using the gray entropy of the entire column,and the result is negative,and the point is constructed as the point.Entropy reduction factor.The result of the entropy reduction coefficient being positive is the point that satisfies the skyline a priori.You can get the skyline by filtering and smoothing the points that satisfy the skyline a priori.The gray entropy is used to construct the entropy reduction coefficient of each point for the column,considering the overall distribution of the image,reducing the influence of local noise,and then obtaining the sparse solution of the skyline before filtering and smoothing.Of course,this method in the image column we get the split point to get more than one split point,but a series of points.The optimal coordinates are obtained by multiple trial adjustments and optimization of the selection strategy,and then the images are segmented and fitted using the selected points.In summary,this thesis will discuss the sky-line detection algorithm based on gray entropy in detail,and compare it with the commonly used skyline detection algorithm;then,further discuss the selection strategy of gray-entropy variant and its fitting function optimization problem,To achieve higher universality.The experimental results show that the detection of non-linear skylines in a complex background,the effect obtained by this method is usually better than other commonly used methods.