Fruit Thermal Imaging Detection Based on Laplacian of Gaussian Algorithm

Traditional fruit grading and damage detection mostly use sensory evaluation methods. With the development of computer vision technology, automatic computer vision detection and grading technology developed rapidly. To solve the problem of fruit damage detection, we propose a technical scheme for fruit thermal imaging damage detection using image processing technology. In this scheme, the Laplacian of Gaussian (LoG) algorithm was used to detect the damaged parts; a Gaussian convolution template is used to suppress noise. Different convolution filter results were obtained by varying the convolution kernel sizes and σ values to enhance the color degree of the damaged part in the image. Then, the edge detection technology was used to obtain the edge information of the damaged part. In the experiment, apples with local damage were selected as the research object, and five evaluation methods, including references and non-references, were selected to analyze the influence of the convolution process on the edge detection of damaged parts. The experimental results show that the LoG algorithm can effectively detect the damaged parts of fruits during thermal imaging, and the influence of the convolution kernel size on the edge detection results is far greater than the value of σ. By increasing the size of the convolution kernel, the edge information of the damaged parts can be effectively deepened. This study provides a feasible solution for fruit damage area detection.