The work aims to design an algorithm and device for automatic counting of compressor slides based on machine vision technology and PLC in view of the cumbersome problem of manually counting compressor slides. Industrial cameras and high-power LED lights were used to design the visual device, and PLC was used as the assembly line controller to communicate with the visual computer through TCP. An automatic counting algorithm was designed based on image processing technology. Firstly, the area of interest was cropped from the boxed slide image collected by the camera and subject to filtering, histogram equalization, morphological processing, and adaptive threshold binary segmentation to enhance image contrast. Then, the image cavities were filled. After the larger noise points were removed, the thin algorithm was used to refine the tightly adhered slides, and the histogram was drawn by the horizontal projection method to find the cutting position. Finally, the connected domain analysis was performed to determine the number of slides, and design the user interface to display the counting results and system status. Through many on-site tests, the overall system ran smoothly and was not affected by external light, with high counting accuracy. Tests were carried out at different time in a day, the average time for each count was 7.6 s, and the accuracy reached 99.83%. The application of the projection method solves the segmentation problem of dense adhesion. This method is highly adaptable to the environment, can automatically count slides stably and accurately and has obvious application advantages in the actual mass production and packaging of slides.