Freezer structure and commodity layout detection method thereof
A surface detection and target detection algorithm technology, applied in image data processing, instrument, character and pattern recognition, etc., can solve the problems of discontinuous horizontal line segments, low detection rate, and inability to estimate empty layers, and achieve accurate statistical results. The effect of improved detection rate, detection rate and accuracy rate
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0043] Such as figure 1 Shown is a flow chart of a preferred embodiment of the present invention.
[0044] Before implementing the detection, first train the freezer door detector and the freezer layer detector based on the method of deep learning. The specific operation steps are as follows:
[0045] 1. Collect the original pictures of various scenes, and try to ensure that the number of photos with the door open is equal to the number of photos with the door closed, such as 500 pictures for the open freezer and 500 pictures for the closed freezer.
[0046] 2. Make a freezer door dataset:
[0047] a. For open-door freezers, mark the cavity of the freezer so that the bounding box just covers it;
[0048] b. For a closed freezer, the freezer door is such that the bounding box just covers it.
[0049] 3. Train the freezer door detector with a deep learning-based target detection algorithm (such as Faster RCNN and its derivatives).
[0050] 4. According to the labeled data, c...
Embodiment 2
[0065] Such as figure 2 Shown is a flow chart of another preferred embodiment of the present invention. In this embodiment, the freezer door detection, freezer layer detection, and commodity detection and identification are all performed on the original image, and then the commodities outside the door and the freezer layer are eliminated. The difference between the alternative method provided by this embodiment and the method provided by Embodiment 1 is that both detection and recognition are performed on the original image, and there is no cropping for recognition, and there are more masking and elimination processes in post-processing. This method only carries out three detections to the original image, compared to the original method for 1+N (number of doors)+M (number of layers) detections (such as a four-layer single-door freezer, N=1, M=4, you need 6 detections), the speed is faster, but because there is no process of cutting and enlarging, the loss of feature details ...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More