Target object detection method and device
A technology of target object and detection method, which is applied in the field of image processing, can solve problems such as difficulty in ensuring the detection accuracy of target objects, and achieve the effects of simplifying the target object detection process, improving detection efficiency, and strong adaptability
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0112] First introduce the offline training process.
[0113] Pre-select several parts that can characterize the target object on the target object. The selected part can be called the setting part. Taking the target object as an example, the selected setting part can be left eye, right eye, mouth, The right side of the chin and the right side of the forehead. The reasons for selecting these five setting components are as follows:
[0114] 1) The eyes and mouth are the most characteristic parts of the human face, which can effectively reduce misjudgments;
[0115] 2) Both eyes have a higher probability of appearing in an image containing a human face, which is more reliable as a basis for judging whether an image contains a human face;
[0116] 3) The combination of the right side of the chin and the right side of the forehead can better demarcate the face area, and the situation that only part of the face exists in the image can be better solved by selecting the same side;
[0117] 4...
Embodiment 2
[0138] The online inspection process is described below.
[0139] Such as figure 2 As shown, it is a flowchart of the target object detection method proposed in the second embodiment of the present invention, and the specific processing flow is as follows:
[0140] Step 21: Perform stepwise zooming processing on the image to be detected according to a preset zoom ratio to obtain a corresponding image pyramid.
[0141] If the preset zoom ratio is step and the size of the image to be detected is psize, the size of the image after zooming once according to the zoom ratio step is psize×step, and the size of the image after zooming twice is psize×(step) 2 , And so on, until the preset zoom times are reached. Among them, the number of zooms can be represented by λ, which represents the number of images included in the image pyramid, that is, the number of layers of the image pyramid.
[0142] The obtained image pyramid is a multi-resolution image structure, and each layer of the image pyra...
Embodiment 3
[0177] Corresponding to the target object detection method proposed in the second embodiment of the present invention, the third embodiment of the present invention also provides a target object detection device, the structure of which is as follows Image 6 Shown, including:
[0178] The zoom processing unit 61 is configured to perform stepwise zoom processing on the image to be detected according to a preset zoom ratio to obtain a corresponding image pyramid;
[0179] The feature determining unit 62 is configured to determine the root HOG feature of each image in the image pyramid according to the root filter corresponding to the target object, and determine the component filters in the image pyramid according to the respective component filters corresponding to each set component in the target object HOG features of parts of each image;
[0180] The window sliding unit 63 is configured to, for each image in the image pyramid, slide the root detection window corresponding to the r...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More 