A Linear Fitting Algorithm Based on Locality Sensitive Hashing with Random Sample Consensus for the Thickness Measurement of Reconstituted Tobacco

For the thickness measurement of reconstituted tobacco by modulated line laser, the linear fitting algorithm directly affects the accuracy of the measurement results. After modulated by reconstituted tobacco, the edges of straight line in the visual image often suffer from noise such as adhesion and spikes, leading to inefficient fitting and poor accuracy. To address this problem, a linear fitting algorithm based on locality sensitive hashing with random sample consensus is proposed. We preprocess the data point set using a locality sensitive hashing (LSH) algorithm that maps similar data points into corresponding hash buckets. Using the random sample consensus (RANSAC) to randomly select sample data points from hash buckets, generate an inlier candidate and fit an initial model. Assess the quality of the model and iteratively optimize it until the specified stopping criteria are met. Finally, repeat the above steps and select the model with the maximum number of inliers. When the residual sum of squares (RSS) is close, the simulation proves that the determination coefficient of the proposed algorithm is closer to 1. The average absolute error of the proposed algorithm is less than the least squares method and RANSAC, etc.